The Effect of Confirmation Bias in Criminal
Investigative Decision Making
Chapter 1: Introduction to the Study
Introduction
In the last half-century, researchers have examined bias and heuristic research
with confirmation bias being investigated by Wasson (1960), useful heuristics explained
by Tversky and Kahneman (1974), and a compendium of confirmation bias research
assembled by Nickerson (1998). More recently, in the last decade there has been an
increased forensic interest in understanding confirmation bias and emotion in legal
settings (Ask & Granhag, 2005; Ask, Rebelius, & Granhag, 2008; Bornstein & Wiener,
2006; Capestany & Harris, 2014; Dror & Fraser-Mackenzie, 2009; O’Brien, 2009;
Rassin, 2008, 2010; Schwind & Budner, 2012; Schrackmann & Oswald, 2014; Snook &
Cullen, 2009). The phenomenon of confirmation bias in criminal investigative decision
making continues to evolve.
Despite the growing interest in heuristics and biases, an area of research that
remains neglected involves practical law enforcement decision making (Wiener,
Bornstein & Voss, 2006). The applicability of these biases under different circumstances
has been the subject of exploration for many different disciplines (Griffin & Tversky,
1992; Kosnik, 2007). Capestany and Harris (2014) reported on the effect of emotion on
legal decision making and the need for recognizing the intention behind the decision
making behavior. West, Toplak, and Stanovich (2008) investigated cognitive biases and
heuristics as aspects of critical thinking. The purpose of this study was to examine the
phenomenon of confirmation bias in context to the nature of law enforcement decision
making in the criminal justice system.
Among the cognition errors that affect information processing and decision
making, confirmation bias has been described as the single most problematic aspect of
human reasoning (Nickerson, 1998). Confirmation bias can lead people to selectively
attend to information consistent with their prior convictions based upon available
information, (Jonas, Schulz-Hardt, Frey, & Thelen, 2001), thereby bolstering a favored
hypothesis (Kosnik, 2008). Confirmation bias may also occur through interpreting
ambiguous evidence in a biased manner or recalling information selectively, often
resulting in biased reframing of information (Dror, 2005). A more detailed discussion on
confirmation bias in decision making is discussed in Chapter 2.
In this chapter, I introduce the psychological phenomenon of confirmation bias
and its potential to adversely affect judgment and decision making in a criminal
investigative setting as the background of the problem. This section will serve as an
overview for this study and will include the problem statement, nature of the study,
research questions and hypotheses, theoretical basis, key definitions of variables and
operational terms, assumptions, limitations, and delimitations. This chapter will also
include the significance of the problem and the social change implications of this study.
Background
In arriving at decisions about evidence, criminal investigators do not always
employ judgment about evidence that is accurate (Dror & Bucht, 2011). Forensic science
pioneer Luke May (1936), noted “the eyes see in things only what they look for, and they
look only for what is already in the mind. Often the most significant piece of evidence is
overlooked or misinterpreted because someone has jumped to a premature conclusion” (p.
59). Technology has changed significantly since May made this observation, but the
social cognition errors made by criminal investigators remain the same. Difficult and
complex circumstances that lead to human error expose the frailty of the decision-making
process when certain information is withheld or mistaken (Manning, 1977). Together the
errors often comprise social cognition errors that can be fatal to determining legal
probable cause (Schrackmann & Oswald, 2014). Therefore, it is not only what the
investigator finds out, but also how they find it that is important.
As human beings, law enforcement officers are susceptible to biases, heuristics,
schema, and emotion that can interfere with information processing and decision making
(Kosnik, 2008; Neuberg & Fiske, 1987). Investigators maintain dual roles as citizens and
as criminal investigative fact finders. They have their own (closed) social world (Klinger,
2004; Manning, 1997), and another, less conspicuous one that they delve into when
dealing with criminal behavior. Both roles develop certain expectations and social
cognition to make enough sense out of their environment and to function adequately
(Manning, 1997). Problems can arise when expectations and environments contaminate
objectivity in decision making. Wilson and Brekke (1994) defined mental contamination
as “unwanted judgment emotion or behavior related to uncontrollable or unconscious
mental processing” (p. 172). By including law enforcement officers as participants in this
research, a more realistic perspective of attitudes and biases may improve on the validity
of confirmation bias research in criminal investigative decision making.
People are generally believers, not disbelievers, and they have problems not
believing something unless they are provided with a reasonable alternative (Douglass,
2000). In a forensic setting, there is an inherent danger when criminal investigators
develop biases, assumptions, or conclusions about evidence, which present inherent
dangers for the wrongly accused and a loss of faith in the law enforcement profession
(Kruger & Dunning, 1999; Taslitz, 2010). The rule of law is an issue of fundamental
fairness and equity in the criminal justice system, and confirmation bias in criminal
investigation can not only lead to wrongful accusations and convictions or denying
people of their rights, but can also by weaken legitimate criminal cases against known
perpetrators (Dror & Bucht, 2011).
In most wrongful conviction cases, bias is a factor present in the form of
subjective judgments that lead to psychological contamination (Dror & Bucht, 2011;
Kassin, Dror, & Kukuka, 2013; Vrij, 2006). A lack of appropriate skepticism toward a
complainant can result in little to no investigative effort to refute a wrongful allegation,
which if confirmation bias took place, could result in a wrongful conviction. Capestany
and Harris (2014) observed that, through brain activity, emotion does inhibit logical
decision making. Whether in the form of a perceived confession, faulty eyewitness
identification, or the problematic use of forensic science, bias is common in wrongful
convictions (Dror & Bucht, 2005; Kassin et al., 2013). This study was needed to examine
confirmation bias from an applied perspective in a forensic setting.
Problem Statement
The problem addressed in this study involves the effect of confirmation bias on
criminal investigative decision making. The main premise of confirmation bias is that
there is a preference for theory consistent information (Ditto & Lopez, 1992; Ditto, 1998;
Gilovich & Griffin, 2002; Kassin et al., 2013; Nickerson, 1998; O’Brien, 2008; Rassin,
2010; Spano, 2005; Tversky & Kahneman, 1974). While it is theorized that confirmation
bias exists in criminal investigative decision making, the subject has not been studied
exclusively in an applied setting. In this study, I attempted to determine whether there is a
preference for confirmatory evidence from a practical law enforcement perspective,
filling an existing research gap.
Emotion can influence an investigator’s belief in the guilt of a suspect in a crime
such as sexual abuse. Experiencing anger can cause the criminal investigator to form a
guilt judgment (Ask & Pina, 2011) or even respond punitively (Capestany & Harris,
2014; Lerner & Tiedens, 2006). When contextual factors combine with an investigator’s
expectations, the officer may distort the interpretation of what he or she or sees (Pronin,
Ross, & Gilovich, 2004), such as when an officer learns that a suspect is a convicted sex
offender. Even when an investigator makes an effort to remain objective, he or she may
revert to a previously held bias or heuristic to make a judgment (Weiner, Bornstein, &
Voss, 2006). When people are aware of their own biases, they are better able to overcome
them (West, Meserve, & Stanovich, 2012). Part of the phenomenon of confirmation bias
that is particularly relevant to this study is that once a conclusion is formed, it is unlikely
to change (Ditto et al., 2002; Nickerson, 1998) even in the face of disconfirmatory
information. Thus, the influence of emotion in the formation of confirmation bias can be
powerful.
Another research gap is the inconsistency of the findings on confirmation bias in
forensic settings. O’Brien (2009) argued that should a preference for incriminating
evidence occur, such confirmation bias can be counteracted if officers think of reasons of
why the suspect may be innocent. Rassin (2008) determined that investigators are prone
to seek evidence of the suspect’s guilt. Further, Rassin concluded that participants did not
favor incriminating over exonerating evidence. Schrackmann and Oswald (2014) argued
that people who are inclined to reach a decision quickly demonstrate rigidity of thought
and reluctance to consider views other than their own. Such rigidity of thought is
inconsistent with best practices in criminal investigation where the investigator must
follow the evidence, even when it excludes a person who may formerly have been a
suspect.
Confirmation bias is generally studied in cognitive and social psychology
laboratories; however, the applied nature of this problem brings it into the public domain
of law enforcement. Existing studies lack an applied component that is critical to
understand confirmation bias in criminal investigative decision making correctly, and
lack any realistic assistance to effect change. In this study, I addressed these limitations to
by incorporating a population of U.S. law enforcement personnel as research participants.
This study built upon the existing research by evaluating various influences on
investigative decision making, such as the type of assignment within law enforcement
(recruit, patrol, investigations), the order in which evidence is presented, and whether
affect can overcome evidence to produce a guilt judgment in an crime with an emotional
element.
Literature Review
There is a small body of research on confirmation bias in criminal investigations
and even less that includes police as research participants. The most salient research on
confirmation bias in criminal investigation that informed the present study was 6 to
8years-old, from studies by O’Brien; (2009), Ask and Granhag, (2005), Ask et al., (2008),
and Rassin (2008, 2010, 2012). Researchers in Sweden (Ask & Granhag, 2005; Ask et al.,
2008) and The Netherlands (Eerland & Rassin, 2012; Rassin, 2008, 2010; Rassin,
Eerland, & Kuijpers, 2010) have studied confirmation bias in forensic decision making in
their respective countries, the results of which were consistent with previous research on
confirmation bias. Rassin, (2010, 2012) attempted to replicate older research (Ask &
Granhag, 2005) and provided support for the concept that people prefer incriminating
versus exonerating evidence when the severity of crime increases.
Ask and Granhag, (2007) conducted research with Swedish police recruits and
examined the elasticity of different kinds of evidence, including photographic, witness
statements, and DNA. Ask and Granhag concluded that the recruits rated disconfirmatory
evidence as less reliable than confirmatory evidence. Rassin, (2010) attempted to
replicate the findings by Ask and Granhag using Dutch police, lawyers, and judges.
Rassin confirmed the confirmation proneness that it had predicted. Rassin investigated
the concept of confirmation bias in criminal investigation as it related to crime severity
and the strength of evidence in a case. Rassin used Dutch law students as participants and
the findings were in line with the predicted hypothesis that criminal procedure inherently
fuels confirmation bias. Rassin also found that the “fact finders” appeared to create
nonexistent evidence of the suspect’s guilt, suggesting that the absence of
disconfirmatory evidence can be interpreted as guilt.
Research by O’Brien (2009) demonstrated that participants displayed a preference
for hypothesis-consistent evidence. In this research O’Brien explored confirmation bias in
the pretext of a law enforcement setting and suggested that the students who considered
alternative hypotheses demonstrated less of a preference for confirmatory evidence.
O’Brien concluded that participants preferred information favoring their hypothesis in a
crime scenario and theorized that remedial measures might counteract this tendency to
prefer theory-consistent evidence including considering counter-hypotheses. Rassin et al.
(2010) disagreed with this assessment, however, and cast doubt on the assumption that
confirmation bias lurks in legal decision making.
Research on the role of mood and emotion in legal decision making is more
current than that of confirmation bias. Capestany and Harris (2014) studied the biological
basis of logical reasoning with functional Magnetic Resonance Imaging (fMRI)
technology. Capestany and Harris examined how more ‘disgusting’ crimes could
influence reason in legal decision making. Busey and Dror (2011) investigated forensic
decision making from an information processing perspective, focusing on sensory
systems. Schrackmann and Oswald (2014) discussed the preference for confirmatory
information when testing attitudes or beliefs, and allowed participants to review further
items of evidence and readdress their level of suspicion toward a suspect. Fahsing and
Ask (2013) concluded that there are certain tipping points in homicide investigations that
are less related to evidence and are more situational and circumstantial in nature, causing
the investigator to begin to build a case rather than investigate it, resulting in excessive
downplaying of conflicting evidence as closed-mindedness. This study attempted to close
the gap between basic scientific findings and the practical nature of legal decision
making.
A significant research gap that emerges in the study of confirmation bias is the
lack of law enforcement personnel as participants when studying confirmation bias.
While some Swedish and Dutch researchers have used law enforcement personnel in their
studies, no U.S. research has included this population in a study. Fahsing and Ask (2013)
included detectives from the United Kingdom and Norway as participants in a qualitative
study on decision making in homicide investigations. Including U.S. law enforcement
personnel in this quantitative study will address the gap in the literature by using certified
police officers representative of the population under consideration: criminal
investigators, patrol officers, and police recruits. I examined the effect of confirmation
bias and emotion on criminal investigative decision making. In addition, I addressed the
gaps in the research concerning the effect of emotion on confirmation in investigative
decision making, as well as the differences between assignments (recruit, patrol,
investigations) and how it can influence confirmation bias.
What research does exist on the subject of confirmation bias in criminal
investigative decision making has borrowed or extended conceptual frameworks from
other social science disciplines and extended them to large convenience samples of young
college students. The findings are based upon generalizations and assumptions about
what data might support, leaving a significant gap in the research (Dabney, Copes,
Tewksbury, & Hawk-Tourtelot, 2013). A more detailed examination of the literature will
be in Chapter 2.
This study will help to identify the issue of confirmation bias in criminal
investigative decision making, and may suggest ways to avoid or minimize it.
Additionally, this study proposes to examine the impact of emotion on legal decision
making to aid in furthering the range of awareness of potential interactions between law
and emotion. These issues are important to any community who is served by a police
department, which includes most all of society.
Nature of the Study
The goal of this quantitative study was to examine the psychological phenomenon
of confirmation bias from the perspective of law enforcement, relative to domain-specific
context and content. The nature of this study was a quasi-experimental, between-subjects,
factorial design. This research design provides the advantage of conducting a study in a
natural setting. A disadvantage, however, is the weakness in internal validity and inability
to infer causation due to a lack of random assignment (Frankfort-Nachmias & Nachmias,
2008). This methodology was chosen to best examine the problem in an applied setting
from a quantitative perspective, with practical research participants.
A high degree of correspondence between the sample, population, and the
sampling frame ensured accuracy in helping to provide a representative sample. Thus, the
sample included experienced criminal investigators who met the parameters of the
research design (Campbell & Stanley, 1963; Frankfort-Nachmias & Nachmias, 2008).
The parameters included certified police officers. It was hypothesized that there may be
different degrees of confirmation bias among the groups; thus, the sample sizes for a
proportionate stratified sample came from three strata: police recruits, patrol officers, and
criminal investigators. A disproportionate sample refers to different sized samples,
generally used to analyze one stratum more intensely (Frankfort-Nachmias & Nachmias,
2008). Criminal investigators were in the larger, disproportionate stratum because of the
objective to evaluate confirmation bias in criminal investigations.
I used the type of assignment within law enforcement (i.e., recruit, patrol,
investigator) as one independent variable, the type of crime (i.e., adult sexual assault or
child sexual assault) as another independent variable, and the order of evidence
presentation (i.e., sequential, simultaneous, or reverse sequential) as a third independent
variable, with a measure of guilt belief (Likert-scaled measure) as a dependent variable.
The ordinal data were evaluated using analysis of variance (ANOVA). Interval data were
collected. The research instruments included Likert-scaled answers that were completed
upon reading a criminal case vignette and evaluating the differences among the
independent variables through the use of ANOVA. This provided the opportunity to
understand the effect of one or more independent variables on the dependent variable,
while controlling for the effects other variables.
The participants for this study were certified police officers in a Midwestern state
attending an accredited, state-sponsored criminal justice training. The participants were
gathered for their mandatory, annual in-service training. The participants included patrol
officers, criminal investigators, and a police recruit class in the basic training academy.
With the permission and consent of the state’s department of criminal justice training
academy, I notified specific classes that had potentially relevant participants (criminal
investigators, patrol officers) of the opportunity to volunteer for the study. Potential
participants heard a brief description of the study followed by an educational brief about
how to participate if interested.
Those officers who volunteered to participate first completed a questionnaire with
some personal and professional demographic information. The officers who volunteered
recorded their responses via a computerized survey. They read vignette information and
complete Likert-scaled questionnaires. Responses were in the form of fixed-alternative
questions on a numbered range of 0 to 10. The participants remained anonymous and
were given a unique code to be used in place of their name to log in and participate in the
study.
This representative sample of law enforcement participants strengthened the
internal validity of the study by reducing extrinsic factors, which may produce differences
between the groups of participants (Frankfort-Nachmias & Nachmias, 2008). Only
participants volunteered and completed a consent form proceeded. Any participants who
did not meet the required status (certified peace officer, investigator or patrol) were not
scored or included in the data.
Purpose of the Study
Wrongful conviction researchers have brought some attention to the problem of
confirmation bias (Dror & Bucht, 2005, Gross & O’Brien, 2008; Simon, 2012; Vrij,
2004), but greater awareness of the effects of confirmation bias on criminal investigative
decision making from a more empirical standpoint can help to bring the issue into
mainstream discussion on policy and procedures with criminal justice administrators
(Maroney, 2006). The purpose of this quantitative study was to effect change through
recognition that confirmation bias is a personal deficiency with no effective way of
changing it. Further, I sought to contribute to the body of knowledge on how to avoid the
mistakes that consistently lead to wrongful convictions and generate more effective
decision making (O’Brien, 2009).
The purpose of study was also to describe how the independent variables (bias and
emotion) can affect the dependent variable (guilt judgment) in such a way that decision
making becomes impaired. By describing what areas impair decision making, policy and
procedure makers can also identify what does not impair decision making and work
toward maintaining objectivity and skepticism through appropriate organizational means.
As Bornstein (2006) noted, an officer’s training and experience could be capable of
offsetting any unwanted effects from a victim’s emotional reactions. Using law
enforcement officers as participants in this study will help to close the gap that exists
from overlooking criminal investigations in the study of bias and emotion in judgment
and decision making.
Research Questions
In this study, I examined the presence of confirmation bias in a criminal
investigative setting. Three research questions guided this study. In particular, this
research was organized around the following research questions and associated null
hypotheses:
RQ 1: Does confirmation bias differ among various assignments (recruit, patrol,
investigators) within law enforcement?
H0 1: Confirmation bias does not differ among assignments.
H1 1: Confirmation bias does differ among assignments.
RQ 2: Can emotionally charged crimes increase belief in a suspect’s guilt? H0
2: Emotionally charged crimes do not increase belief in a suspect’s
guilt.
H1 2: Emotionally charged crimes can increase belief in a suspect’s guilt.
RQ 3: Does the order of evidence presentation influence belief in a suspect’s
guilt?
H0 3: Evidence presentation order does not influence guilt belief.
H13: Evidence presentation order does influence guilt belief.
To examine the phenomenon, the dependent variable of confirmation bias was
represented as a confidence assessment of guilt, or guilt belief, relative to the independent
variables (e.g., duty assignment, scenario condition, evidence presentation order) with
which they interact. Variables are explained in greater detail in Chapter 3.
Theoretical and Conceptual Support
There is a lack of theoretical specificity surrounding the phenomenon of
confirmation bias in an applied criminal investigative context. Nevertheless, some
theoretical support for this study emerged from expected utility, rational choice, and
prospect theory. Conceptual framework of the study is further supported through the
application of bounded rationality and heuristic and biases processing. Chapter 2 will
provide a more detailed evaluation of the literature.
Expected Utility
Expected utility helps to explain how people make decisions when uncertainty is a
factor. Originally an economic model that helped explain decisions in risky situations,
expected utility has proved to be relevant in any area of human existence where a choice
with risk is considered (Fiske & Taylor, 2013). Expected utility holds that the intensity of
the outcome is as serious as the circumstances precipitating the judgment (Fiske &
Taylor, 2013). As social perceivers, people are constantly receiving information from
their environment and making choices from the most mundane to the most serious. The
volume of material could occupy all of a person’s time just to get through a day, so
people employ strategies to save time as they make judgments. Expected utility accounts
for many decisions of choice where risk is involved, however slight or serious. Expected
utility theorists maintain that each choice available as an alternative for people includes a
designated value and a probability of occurrence (Fiske & Taylor, 2013). People will
assess those choices for the likelihood and worth of the outcomes and calculate the utility
of each outcome, choosing the one that optimizes their utility.
The theory of expected utility is helpful when examining different outcomes, but
it does not translate fully to human decision making and not to the criminal justice field
where subjective probabilities are not acceptable. When considering criminal choice,
attitudes toward risk are central. Criminal choice does not allow for various transitive
preferences to be known to the decision maker (Fiske & Taylor, 2013). Further, the effect
of context on decisions and preferences, along with the interaction of the effect of context
on decisions and different levels of probability, cannot be explained by expected utility
theory or real world consequences (Fiske & Taylor, 2013). In some instances, the best
choice might be no choice at all. Consequently, a limitation associated with this theory is
the position of assuming that people are rational, logical, and systematic thinkers. A more
accurate understanding about the psychology of problem representation is necessary.
Rational Choice
According to the theory of rational choice, patterns of behavior in society reflect
the choices made by individuals as they try to maximize their benefits and minimize their
costs (Griffin, Gilovich, & Kahneman, 2002). In the classical model of rational choice, a
rational actor chooses what options to pursue by assessing the probability of each possible
outcome, discerning the utility to be derived from each and combining the two
assessments. A rational actor will choose options by assessing the probability of each
outcome compared to the potential utility to be gained, combine the two assessments, and
pursue the option with the greatest combination of utility and probability (Tversky &
Kahneman, 1986). The outcome or potential utility can be recognized as the level of
belief in a suspect’s of guilt and subsequent investigative action. These theories extend to
criminal investigative decision making by examining outcomes (i.e., belief in guilt) and
maximizing the effects (i.e., strength of evidence).
Prospect Theory
Prospect theorists claim that people select alternative choices by assessing risk
where the probabilities are known. Effect, which is a tendency to underweigh outcomes
that are merely probable, in comparison to outcomes obtained with certainty (Kahneman
& Tversky, 1979). According to prospect theory, people’s ability to comprehend extreme
probabilities is limited, meaning events that are highly unlikely are either ignored or
undervalued (Tversky & Kahneman, 1979). This can limit the consideration of a full
range of probabilities. Prospect theorists consider perception and reasoning; the decision
maker weighs probable outcomes with those that are certain. This means that decisional
weights will measure the impact of an event on the desirability of its prospects, not just
on their perceived likelihood. By focusing on the perceived likelihood, the perception
may be subject to biases, as well as other considerations including ambiguity or
vagueness.
Bounded Rationality
This study drew upon the psychological framework of Simon (1955, 1971, 1979).
Simon proposed a more limited criterion for actual performance. Bound, refers to
something being limited and rationality, in this context, refers to normative standards.
Thus bounded rationality refers to decision making limited to the information that a
person possesses, along with limited cognitive ability and amount of time they have to
make a decision. Simon (1955) found that people will avoid the consideration that they
could be wrong, similar to the concept of confirmation bias (Nickerson, 1998). The
concept of bounded rationality revises many of the assumptions of rational choice theory,
as perfectly rational decisions are often not feasible due to the limited resources available
to make them.
This framework was relevant to the present study, as a criminal investigator will
have limited time and information on which he or she may deliberate and is often
required to make a decision. According to this framework, even when a person is
confronted with facts that contradict his or her thoughts, he or she may simply will a way
to perceive things otherwise, thereby maintaining their beliefs. The criminal investigator
may also be confronted with an ipsative situation, where no decision is not an option,
such as when evidence is obvious but lacks context or meaning to the investigator, and it
is not immediately clear what to do with the evidence. Even though an optimal decision is
preferable, a satisfactory one may be all that an individual can accomplish (Erzinclioglu,
3002; Innes, 2003). For example, short-lived (e.g., blood, fluid) evidence may disappear
quickly and a decision must be made whether or not to expend the time and effort
necessary to secure it properly. The evidence, however, may not have an obvious
connection to the case and could be a waste of limited resources.
Because a decision maker lacks the ability and resources to arrive at an optimal
solution, they may instead simplify their choice and then apply their rationality
accordingly. This type of decision making is satisficing, or settling for a satisfactory
solution, rather than searching for an optimal one (Gigerenzer, 2010). Rationality, in this
sense, may come in the form of an emotion such as anger or sympathy and is based
largely on heuristics. In the present study, it was theorized that emotion might play a part
in decision making when confronted with limited information. Out of anger at the
offender, and in sympathy for the victim, the investigator may be willing to satisfice,
settling for enough evidence rather than seeking all of it.
People generally reason and choose rationally, but only within the constraints
imposed by their limited search and computation abilities. Because people have different
levels of ability and effort to perceive, organize, and process information, they may not
arrive at the same conclusion, even though they may have few choices (Gigerenzer &
Selton, 2001). Bounded rationality helps to describe how a person’s belief may be biased
through the limited information they possess or by their beliefs. According to this
framework, even when a person is confronted with facts that contradict their thoughts,
they may simply will a way to perceive things in such a way that they maintain their
beliefs. Even when forced, they tend to concede less than the facts that contradict their
position might otherwise support (Weiner et al, 2006). As investigators use heuristics to
make decisions, bounded rationality must considered in the context of investigative
decision making and the demands of the police environment (Snook & Cullen, 2009).
In addition to bounded rationality, the psychological frameworks of heuristics and
biases (Kahneman & Tversky, 1982), information processing, and anger attributional
affects (Quigley & Tedeshi, 1996) are influential in forensic decision making and were
used help to understand the current study about how law enforcement decisions are
impacted by external forces with psychological explanations.
Definitions
The following list of terms provides relevant definitions pertaining to this research. Other
definitions may exist but may not represent the intended use in this study.
Actor/observer affect: This occurs when a person relates another’s behavior to
dispositional factors but relates his or her own behavior to situational factors (Choi &
Nisbett, 1988).
Affect: Affect refers to a subjective feeling or attitude. Affect occurs through the
automatic, subconscious processing of an emotional feeling or experience and is viewed
with a positive or negative valence (Lazarus, 1982).
Attribution theory: This concerns how people infer causal relations and the
dispositional characteristics of others (Taylor & Fiske, 2013).
Assessment of confidence: For this study a confidence assessment refers to a
degree of guilt belief (see guilt judgment).
Belief perseverance: This refers to a belief that perseveres even when the original
evidence it was based upon is refuted or discredited (Ross, Lepper, & Hubbard, 1975).
Belief Scale: A scale of belief in a person’s guilt relevant to a specific crime
ranging from not guilty at all to definitely guilty. Within those parameters exist
reasonable suspicion, preponderance of evidence, clear and convincing evidence, and
beyond a reasonable doubt; evidence of guilt from least to most, respectively.
Beyond a reasonable doubt: This is the legal threshold required for a criminal
conviction. It is greater than a preponderance of evidence, but less than beyond all doubt.
This is the highest form of legal proof required by the law (Black & Nolan, 1993).
Bias: Refers to a preference for one thing over another.
Clear and convincing evidence: A legal threshold that is greater than a
preponderance of evidence, but less than beyond a reasonable doubt (Black & Nolan,
1993).
Cognitive load: This refers to the number of demands simultaneously made upon
the human mind, potentially impairing accuracy. Limits on cognitive load may also
impair accuracy (Sternberg, 2008).
Confidence assessment: A belief in guilt toward a specific suspect, concerning a
specific allegation. This assessment includes a Likert-scaled judgment of guilt from 0 (not
guilty at all) to 10 (definitely guilty).
Confirmation bias: Confirmation bias in a criminal investigative setting may come
in the form of focusing exclusively on one theory of a case while failing to consider
others for any number of reasons. It may occur by searching for evidence to support the
singular theory, thereby ignoring any other evidence that may be relevant, or
disconfirmatory. It also applies to specific items of evidence such as the veracity of a
witness statement or the importance of a piece of physical evidence (Nickerson, 1998)
Criminal investigation: The process of legally inquiring and gathering evidence of
a crime that has been committed (Lyman, 2011)
Decision making: Refers to the actions made by a police officer during a criminal
investigation.
Dissonance theory: Is concerned with selective learning and attention of
attituderelevant information. The hypothesis of selective perception to support a person’s
attitudes include selective exposure, selective attention, and selective interpretation.
Although evidence for selective exposure entails mostly de facto rather than deliberate
evidence, evidence for selective attention and interpretation supports the premises of
dissonance theory (Taylor & Fiske, 2013).
Emotionally charged crime: This refers to a crime that might shock the conscience
of a citizen and may still cause the police officer to be emotionally aroused (such as a
child sexual assault) whereas a crime of routine occurrence may not.
Exculpatory: Evidence or information that tends to show innocence of a crime
(Black & Nolan, 1993).
Emotions: The compilation of miscellaneous feelings, mood, and affect from
which a state of mind emerges (Maroney, 2006).
Extraordinary emotional provocation: This refers to the reporting of a crime of
such nature that it has the potential to provoke strong emotion. In this context, it must be
recognized that in the field of law enforcement; reported crimes like this are a regular
occurrence.
Exculpatory: Evidence that tends to show innocence (Black & Nolan, 1993).
Feelings: Feelings are conscious experience of happiness, sadness, compassion,
hate and so on; however, feelings are distinctly different from emotions. Emotions may
emerge from feelings (Taylor & Fiske, 2013)
Fundamental attribution error: The tendency, especially in U.S. culture, to judge
an individual's actions as stemming from fundamental personality traits rather than from
the situation in which he or she finds him or herself (Snook & Cullen, 2009). People are
willing to make quick and confident judgments of a subject's personality trait based on a
limited data sample. They will also over-generalize, treating one perceived negative
personality trait as indicative of an overall negative personality across many criteria
(Fiske & Taylor, 2013).
Guilt judgment: Also referred to as a guilt belief or confidence assessment, in this
study a guilt judgment refers to the degree to which an officer believes a suspect to be
guilty, either in the form of a scale (1-10) or a yes/no judgment. A degree of belief is also
referred to as an assessment of confidence (Griffin & Tversky, 1992).
Heuristics: These are mental shortcuts that help people reduce everyday complex
problem solving into simpler judgmental operations in order to meet various
environmental demands (Kunda, 1999). They are often effective strategies that provide
efficient for handling complex data and then drawing conclusions from that information
(Snook & Cullen, 2009). Not all heuristics are the same. Some are useful, simple, and
commonly employed, while others are more sophisticated and idiosyncratic. Tversky and
Kahneman (1974) and Nickerson (1998) focused on some of the most commonly used
heuristics: availability, representative, and anchoring and adjustment.
Illusory correlation: Imposes a relationship where none actually exists (Kunda,
1999).
Inculpatory: Evidence or information that tends to show guilt of a crime (Black &
Nolan, 1993).
Optimizer: A person who holds out for the best possible inference or decision
(Gigerenzer, 2010).
Preponderance of evidence: Also known as the balance of probabilities, the
preponderance of evidence means legal proof by information that, when compared to the
information opposing it, leads to the conclusion that the fact at issue is more probably
true than not (Black & Nolan, 1993). The preponderance of evidence is used in all civil
cases because if the standard is met, the proposition is more likely to be true than not.
This standard is also described as greater than 50% of the proof, more probable than not,
and more likely than not (Federal Rules of Evidence).
Probable cause: Probable cause is a standard used in justifying various police
actions such as searching and arresting. It consists of more than mere suspicion, but less
than the amount of evidence required for conviction (Black & Nolan, 1993).
Reasonable suspicion: Has been defined by the U.S. Supreme Court as “the sort
of common-sense conclusion about human behavior upon which practical people...are
entitled to rely.” It is more than a “hunch” or “gut feeling”. It requires articulable facts or
circumstances that give rise to more than a bare, imaginary, or purely conjectural
suspicion (Black & Nolan, 1993).
Salience. Refers to when a person notices something and devotes attention to it,
and it becomes more prominent and available in the mind and can exert greater influence
on subsequent judgments (Fiske & Taylor 2013).
Satisficing: Making (or settling for) adequate inferences or judgment rather than
an optimal one (Gigerenzer, 2010).
Social cognition: How people make sense of themselves and others in their
environment (Fiske & Taylor, 2013).
Self-serving bias: This is the tendency to take credit for success, and deny
responsibility for failure attributions of responsibility. This also manifests as the tendency
to evaluate ambiguous information in a way that is beneficial to a person’s own interests
(Fiske & Taylor, 2013).
Situation awareness: This refers to an internal construal of the situation on which
a decision is to be based. This process consists of interpreting and combining information
in a specific environment in order to understand what is happening and why.
Assumptions
A study with law enforcement participants requires several basic assumptions.
One assumption was that the duty of criminal investigation falls to law enforcement and
that most major crimes are handled by officers with specialized training who work as
criminal investigators. It was assumed that investigative decisions are based upon
probable cause and reasonable articulable suspicion. It was further assumed that all law
enforcement officers endeavor to behave ethically, professionally, and with a duty to
serve the public.
The most significant assumption in this study was that confirmation bias is a
ubiquitous phenomenon in human existence (Nickerson, 1998) and should be expected to
manifest itself in nearly every area of human judgment (Fiske & Taylor, 2013, Tversky &
Kahneman, 1973, 1974). Another main assumption of this study was that certain features
of this research such as confirmation bias, judgment, emotion, and decision making can
be defined using specific criteria. Another main assumption was that the tasks in this
study approximate real world criminal investigative tasks, crime scenarios, and evidence.
Limitations
The unique nature of prosecution, and by extension criminal investigation,
requires a certain prejudging of guilt. The duty to do justice requires scrutinizing the
evidence against the defendant, not merely for sufficient proof to a guilty verdict, but for
proof that persuades the prosecutor of the defendant‘s guilt beyond a reasonable doubt in
their own mind. Investigative theories are generated based upon this scrutiny, thus, a
necessary measure of bias automatically exists in early stages of criminal investigation by
design.
Another limitation of this study was that it is impossible to know for certain what
variables may or may not have influenced the decision maker and their private
perspectives. Further, if initial steps are determined to be confirmatory, an entire
investigation is not necessarily biased or flawed. Conclusions must be understood in
context to the reality of the dynamic nature of criminal investigations and evidence. There
is a certain amount of subjectivity involved in criminal investigation that will never be
eradicated and would be impossible to quantify
One problem associated with the correlational coefficient in this study was that the
findings about attitudes toward sex offender suspects are measured largely against
morals, societal norms, and a basic sense of fairness. A more empirical measure of
validity in a legal context is the whether or not a person is afforded due process, a clause
within the 14th amendment of the U.S. Constitution enumerating that no person shall be
denied life, liberty, or property without due process of law. Another measure is proof of
beyond a reasonable doubt, which is the standard for a conviction required in criminal
cases (Meyer & Weaver, 2006).
Significance
Approaching a criminal investigative situation with a presumption of guilt (or any
presumption at all) may cause an investigator to be insensitive to potentially exonerating
information, at a minimum. All wrongful convictions begin with a breakdown in the
investigative or adjudicative processes (Simon, 2012). This is especially true when
dealing with human testimony, which may be faulty (Ask & Granhag, 2007; Kasin, Dror,
& Kukucka, 2013; Vrij, 2009). Thus, the danger in approaching cases in this manner is
the possibility of a wrongful conviction, affecting the whole of society. Accordingly, it is
necessary to experiment with the closest possible participants to the relevant influences in
order to capture the dynamic nature of law enforcement decision making.
Awareness and education may be effective tools in combating bias in decision
making, and it is important for forensic psychology professionals to understand this in
legal settings. This study may aid in addressing influences toward confirmatory thinking
in an investigator’s work environment and may be able to increase objective investigative
methods. It is conceivable that policies and procedures in law enforcement can be
adjusted to decrease bias and make more effective decisions. The potential for this area of
study is vast.
The psychological understanding of criminal investigative bias can illuminate the
motivational process of social cognition in investigative decision making. The findings of
this study can be used to examine legal judgments, policies, and procedures. It is
conceivable that psychologists can impact policy making to incorporate mechanisms to
mediate social cognition errors in criminal investigative decision making. Armed with
knowledge of how confirmation bias influences investigative decision making in a
forensic setting, psychologists can also have an impact on related forensic issues such as
interviews and interrogations, eyewitness testimony, and line-up/show-up suspect
identification, all of which are all subject to human judgment and can benefit from an
understanding in this area. These are benefits not only law enforcement and the criminal
justice system, but also the whole of society.
Summary
In this research, I studied bias and emotion that can interfere with criminal
investigative decision making. Emotion may be a moderating variable in investigative
decision making. It is theorized that emotional, evaluative judgments are hard not to make
and angry attitudes toward offenders may remain. Further, there may be certain
moderating effects on social cognition such as prior felony arrests, felony convictions,
sex offenses, or violent offenses that are known to an investigator and can cause him or
her to be more suspicions or willing to make a determination of guilt prematurely.
Anger/emotion may transcend situations and increase different attributions that produce
and stimulate varying levels of information processing, which are at times irrelevant to
the anger-provoking event.
In Chapter 2, I will discuss the relevant literature surrounding the phenomenon of
confirmation bias and emotion in decision making within the forensic realm and will also
explain gaps that exist in the research and how the theories of rational choice and
expected utility support the phenomenon.
Chapter 2: Literature Review
Introduction
The deleterious effects of confirmation bias on criminal investigative decision
making exists on a continuum of behavior ranging from major miscarriages of justice
from wrongful convictions or accusations to the acquittal a guilty defendant. Neither
outcome forestalls the guilty offender from reoffending undeterred and eradicating such
bias is clearly beyond any human effort; however, recognizing common features that
identify and attenuate confirmation bias in a criminal investigative setting is a step in the
right direction. This study, shall examine the elements of confirmation bias and emotion
within the framework of social cognition in criminal investigation.
Human rationality is limited; unfortunately irrationality has no constraints. The
acquisition of intelligence does not automatically bestow competence on a person and
people frequently fail to employ intellectual competence in their decision-making (Kruger
& Dunning, 1999). Because criminal investigators generally work backwards from the
scene to a crime’s origin with limited information, it is incumbent upon them to withhold
judgment until they have gathered the last piece of evidence in a case, lest they arrive at a
premature conclusion. Nevertheless, as human beings, investigators are susceptible to
errors in social cognition, therefore, a unique problem for psychology and law is to
identify the impetus that produces an error in social cognition at critical moments, and
results in a deleterious effect in criminal investigative decision-making.
Decision making and information processing abilities are often not optimal due to
the complexity of information confronting the criminal investigator (Dror & Bucht,
2011), creating the likelihood of overwhelming their cognitive abilities and forcing the
investigator to rely upon available heuristics and perception (Bornstein & Nemeth, 1999).
Since criminal investigations typically operate in reverse linear fashion, they require the
investigation to work backwards from the crime scene to the identification of a
perpetrator, collecting evidence along the way (Lyman, 2011). Unique situational and
contextual pressures further exacerbate the complexities of an investigation such as the
seriousness of a crime, the dynamism and ambiguity of evidence, time constraints
(Rassin, 2008), institutional pressures, the need for closure (Ask & Granhag, 2005), and
others such as overconfidence or even incompetence (Kebbell, Muller & Martin, 2010).
For this study, confirmation bias is best understood as manifesting in three main
ways, each emerging differently: the tendency to seek only that evidence that supports a
favored theory or hypothesis, the rejection of disconfirmatory evidence, and the tendency
to reconstitute vague or ambiguous evidence in such a way that it supports an initial
belief. Once an investigator arrives at a conclusion prematurely, confirmation bias leads
them to maintain their belief, often in light of disconfirmatory evidence (Hernandez &
Preston, 2013; Jonas, Schulz-Hardt, Frey & Thelen, 2001; Tversky & Kahneman, 1974).
Investigators are often guided by their initial theories or hypotheses in their search for
evidence (Ask & Granhag, 2005) and ambiguous evidence is often interpreted as
supporting their expectations (Nickerson, 1997; O’Brien, 2009; Spano, 2005). As a result,
confirmation bias can induce the investigator to ask only those questions that are likely to
confirm their hypothesis and commit to one single causal explanation (Ask & Granhag,
2005). Once a conclusion is reached it is cognitively adopted as factual (Dror &
FraserMackenzie, 2005), and once adopted as factual it is not likely to be amended.
People are as motivated to find correct beliefs and decisions as they are to defend those
they have already formed (Johnson & Fuita, 2012). With this approach, the tendentious
investigator transforms from fact gathering to case building.
The foundation of the phenomenon of confirmation bias in modern research relies
heavily upon the seminal research on heuristics and biases of Tversky and Kahneman
(1971, 1973, 1974, 1981, 1983, 1986, 2002; see also Kahneman & Tversky, 1972, 1973,
1979, 1986, 1996, 2000; Kahneman, Slovic, & Tversky, 1982), and meta-analysis of
confirmation bias by Nickerson (1998). These studies spawned dozens of subsequent
studies, however, none have addressed the practical aspect of confirmation bias or
emotion in criminal investigative decision making.
Several different aspects of confirmation bias in the research pertain to legal
decision making. Ask, Rebelius, and Granhag (2008) referred to the concept that scrutiny
toward personal interpretation tends to be with positive implications for an established
personal belief or value rather than negative implications as “asymmetrical symmetry”.
Nickerson (1998) isolated the intentionality of confirmation bias as “building a case to
justify a conclusion that is already drawn” (p. 175). Schrackmann and Oswald (2014)
described biased searching or evaluation of information as a process that “either
systematically enhances the probability of the corroboration of a formed belief, or
impedes its possible refutation” (p. 5). Fahsing and Ask (2013) noted that confirmation
bias includes ignoring or downplaying material that conflicts with the available evidence.
Faulty interpretations of evidence can lead to overconfidence in a belief of guilt and often
contribute to wrongful convictions (Dror & Bucht, 2005; Griffin & Tversky, 1992; Gross &
O’Brien, 2008; Huff, 2004; Simon, 2012). A key distinction of this phenomenon is that some
belief exists with no clear evidence to support the interpretation (Klayman & Ha, 1987; Kosnik,
2007; Oswald & Grosjean 2004).
Another issue associated with confirmation bias is emotional decision making.
When confronted by social stimuli, a law enforcement officer’s reaction may be
automatic, habitual, unthinking (Kahneman & Tversky, 1986; Kahneman, 2011),
intuitive, or deliberative (Kruglanski & Gigerenzer, 2011). Emotions are known to
influence fact finders’ attributions of responsibility (Bornstein & Wiener, 2006). Emotion
that precedes decision making in criminal investigative contexts has the potential to begin
a process that is flawed from the beginning, since bias, by its very nature, can
selfperpetuate (Capestany & Harris, 2014; Johnson & Fujita, 2012). Additionally, people
largely fail to detect evidence of their own bias (Pronin, Kruger, Savitsky, & Ross, 2001)
and when they look for it, they are apt to conclude that none exists (Kahneman, 2011).
Emotion has the potential to be a significant problem in criminal investigative decision
making.
Emotional contexts that precede legal decision making can override rational
thinking and decision making (Bornstein & Wiener, 2006; Shiv & Fedorikhin, 2002).
Emotional decision making can be problematic if personal perception overcomes
objective evidentiary evaluation and a subjective threshold of guilt is considered
(Feigenson & Park, 2006). The belief in one’s own accuracy can lead to the belief that
their perceptions reflect the true nature of things (Pronin, Kruger, Savitsky, & Ross,
2001). If emotion can cause evidence to be perceived through a new framework, it can
lead to different perceptions, distortions, and conclusions (Feigenson & Park, 2006).
Evidence must be explained as it is, rather than as it could be, for if our belief correlates
with our biased perception, as West, Meserve, and Stanovich (2012) observed, the latter
will dictate the former.
A core problem with confirmation bias is that it is pervasive and perpetual in the
human experience (Johnson & Fujita, 2012). In the process of making decisions criminal
investigators are subject to the same kinds of errors as lay persons (Innes, 2003). This
study proposes to investigate the phenomenon in a criminal justice setting with law
enforcement participants from different levels of experience, from basic police recruits to
experienced criminal investigators. This research will explore whether or not the criminal
investigator demonstrates a preference for confirmatory evidence and what variables, if
any, correlate with objectivity in criminal investigative decision making. This study will
also examine the effect of emotion on investigative decision making. In addition, this
study will explore the relationship between confirmation bias and criminal investigative
decision making, and its effect on guilt judgment.
The last decade has witnessed a burgeoning interest on confirmation bias literature
with a forensic application of legal decision making. Swedish studies (Ask & Granhag,
2005; Ask, Rebelius, & Granhag, 2008), Dutch studies (Eerland & Rassin,
2012; Rassin, 2008, 2010; Rassin, Eerland, & Kuijpers, 2010), and an Australian study
(Wastell, Weeks, Wearing, & Duncan, 2012) have explored the issue in a forensic setting,
while O’Brien (2009) conducted the lone identifiable study on confirmation bias in
criminal investigation in the US. Other cognitive and psychological phenomenon that
confirm the current study include research on perception, judgment and decision making
(Dror & Fraser-Mackenzie, 2009; Gilovich, Griffin, & Kahneman, 1998; Kahneman,
Slovic & Tversky, 1982; Masnick & Zimmerman, 2009; Nickerson, 1998; Nisbett &
Ross, 1980; Rossmo, 2009; Snook & Cullen, 2009).
It is well documented that criminal investigators have a tendency to perceive
suspects as very likely to be guilty (Baldwin, 1993; Leo, 1996; Meissner & Kassin,
2002). Anger is known to increase attributions of blame and intent to people who cause
negative events (Goldberg, et al., 1999; Quigley & Tedeschi, 1996; Young, Tiedens,
Jung, & Tsai, 2011). In the present study it is hypothesized that vivid details that inflame
emotions can spark anger in the investigator, who may then focus on the factors
incidental to the evidence (e.g., learning a suspect is a registered sex offender may
confirm an investigator’s belief in guilt). Research by Gollwitzer (1990) showed that
when an investigative mindset focuses on specific course of action, a more deliberative
mindset is assumed, suggesting that an investigator with a deliberative mindset may
search only for confirmatory evidence.
Capestany and Harris (2014), Fahsing and Ask (2011), and Maroney (2006)
confirm the problem of affective responses in investigative decision making. Quigley
(1996) observed that anger led to stronger attributions of blame, while Ask and Granhag
(2007) studied how anger toward sex offenders can produce strong affective reactions,
supporting the hypothesis that there is a relationship between the type of offense and the
strength of a guilt judgment by an officer. Cognitively, Sternberg (2006) describes the
heuristic of confirmation bias as an objective to authenticate a criminal investigator’s
beliefs, influencing both reasoning and decision making. Ruscio (2006) describes
confirmation bias as detrimental to evidence-based reasoning and efficient decision
making. As Bornstein and Nemeth (1999) noted, it is rare that a person can ignore their
emotions in a criminal case.
This review begins with the relevant theories associated with the phenomenon,
and will continue by discussing the effect of this bias on guilt judgment by the criminal
investigator, and how it can impact a crime scenario. This literature review will include
comparisons of former studies and the gaps associated with each study that have led to
the need of this research study. Further, this literature review will investigate the
relationship between confirmation bias and criminal investigation, examining such
variables as the type of duty assignment and the type of crime, as well as the relationship
between demographic information and the tendency to engage in biased decision making.
What follows is how the search was conducted and how the literature was identified,
organized, and synthesized.
The literature review starts with discussion of decision making, social cognition,
heuristics and biases, information processing, in decision making. The review then turns
to several influential theories in social psychology - prospect theory, rational choice and
expected utility, and frameworks that explain the phenomenon such as bounded
rationality and heuristic and biases study. The review concludes with a section on affect
and decision making.
Literature Search Strategy
This literature review began with a search of the library databases within the
Walden Library in both psychological and criminal justice databases. A computerized
search was conducted to identify articles regarding the concepts of confirmation bias,
decision making, reasoning, judgment, intuition, legal and criminal investigative decision
making, and for material that intersected the domains of forensic psychology and social
cognition. Searches were initially set to recognize studies from the years 2010 to 2014 for
current research and a comprehensive search examined the years from 1970 to 2013.
The review was conducted via tertiary Libraries with electronic EBSCO
databases, mostly utilizing psychological databases including psycINFO,
psycARTICLES, ProQuest, Google Scholar, Psychology: SAGE Full-Text, SocINDEX,
ERIC and Google Scholar. Criminal justice databases included Lexis-Nexis Academic,
Legal Trac, Criminal Justice Periodicals, Oxford Bibliographies Online: Criminology,
SocINDEX, ProQuest Central, and Google Scholar.
Using the search terms bias AND investigations in ProQuest Criminal Justice
returns 66 results. The virtual Boolean search parameters that were utilized included
multiple combinations of keywords including confirmation, bias, decision making,
perception, criminal, investigation, investigations and investigative, and included
variations of keywords.
The phrase confirmation bias produced 133 results, however, separating the terms
as Boolean parameters produced 398 results, 45 of them books, 39 dissertations, 11 book
reviews, and the remaining 308 journal articles. Of the journal articles, 306 were
peerreviewed. Using the terms criminal AND bias returned 219 results. Among these
results are articles on confession evidence, hate crime, gender bias, hypnosis, and racial
bias, economic bias, interview, lineup, jury selection, and a few on confirmation bias in a
criminal investigative setting.
Multiple combinations of search terms were conducted in different library
databases. In ProQuest Criminal Justice database, search terms investigator AND bias
returns 19 results, investigative AND bias returned three results, confirmation AND bias
return eight results, and confirmation AND bias AND criminal returns two results, none
of which are related to criminal investigative bias. Using search terms decision AND
making returns 22,870 results. In PsycINFO, using Boolean search parameters
confirmation AND bias returned 133 results. Among the results were several articles of
relevance. Adding the third search parameter investigation yielded 10 results, three of
which were specifically relevant to this study. The search terms confirmation AND
investigative AND bias returned eight results, which included articles already gathered
for this study. Using criminal AND bias returned 386 results, many of which were on
racial bias as well as lineup, gender, examiner, interview and organizational biases.
Among the results are articles on confession evidence, hate crime, gender bias,
hypnosis, and racial bias, but little related to confirmation bias in a criminal investigative
setting. Using search terms bias AND investigations in ProQuest Criminal Justice search
turns up 66 results Using the same search terms in different psychology and criminal
justice databases provided no further results. The phrase investigator bias produces
results that focus on the detection of deception in interview settings and false confessions.
This body of research discusses the concept of interrogation strategies that potentially
lead to statements by the accused. While this information has relevance to false
confession, it does not consider the role of the criminal investigator’s judgmental process,
perception, cognition and practical aspects of the role of the criminal investigator.
Sources were also located from literature reviews and meta-analyses that
addressed confirmation bias and bias in criminal investigation. The type of literature
searched included the domains of cognitive neuroscience, cognitive and social
psychology, information processing, heuristics and biases, decision making and
judgment, criminal investigative decision making, critical thinking, and criminal justice.
Contained within the body of research related to bias is literature related to interview and
interrogation, racial profiling, criminal profiling, lineups and show-ups, and eyewitness
reliability. As they are related to legal problems, they are often framed in the context of
forensic bias, however, while forensic in nature, these issues are only tangentially related
to confirmation bias in the context of criminal investigative decision making.
There is approximately 40 years of heuristics and biases research in cognitive and
social psychology literature (Evans, 1989; Kahneman, 1973, 2003, 2011; Kahneman &
Frederick, 2002; Kahneman & Tversky, 1972, 1973, 1996, 2000; Tversky & Kahneman,
1971, 1973, 1974, 1981, 1983, 1986, 2002; Wason, 1966). A keyword search in Google
Scholar produces 419,000 results with Nickerson (1998) listed as the first response.
Nickerson (1998) also leads other articles, having been cited 1139 times. Among the most
commonly cited articles in the literature on the domain of biases and heuristics included
Wason (1960), Simon (1959; 1971), and Tversky and Kahneman (1973; 1974;
1979; 1986; 2000).
Theoretical Foundation
For a theory to be relevant to criminal investigation there must be qualities to
account for objectivity. Since this disqualifies many theoretical frameworks, this study
shall focus on those theories and frameworks that lend themselves to objectivity in
decision making. There are several theoretical concepts combine to explain the concept of
bias and emotion in criminal investigative decision making, including information
processing, heuristics and biases and affect.
Rational Choice
The study of modern day heuristic and bias research emerged from the model of
rational choice. It is generally accepted as rational that we want more of something good,
rather than less of it. This simple explanation of behavior found its way into nearly all
modern economic textbooks and other realms of human decision making. The framework
of rational choice allows for an understanding of both economic and social behavior. In
the classical model of rational choice, a rational actor chooses what options to pursue by
assessing the probability of each possible outcome, discerning the utility to be derived
from each, and combining the two assessments. One assumption of this theory is that the
human being is a rational actor and will pursue the option that offers the optimal
combination of probability and utility.
According to rational choice theory, decisions determine behavior and follow a set
of general laws (Satz & Ferejohn, 1994). This theory generally explains a person’s
actions in terms of her or his own mental state. In the present study this can be understood
as looking at certain preferences of criminals. A child sex offender, for example, has very
specific preferences that investigators recognize and they may come to expect consistency
from the child sex offender.
Expected Utility
The theory of expected utility assumes that the decision maker has complete
control over all of the potential outcomes that might occur based upon their decision
(Baron, 2007). Although we may have defined a continuum of good to bad, we all have
deviations from that continuum from which our individual biases are conceived in the
form of tastes, desires, and expectations that we develop through our life experiences.
(Caplin & Lehey, 2001) explained expected utility theory to anticipated emotions. This
theory applies to the present study in this limited sense, for the purpose of recognizing the
need for consistency in behaviors. Expected utility applies to the present study in that
people adopt decision making strategies to accomplish broad objectives to reduce risk. In
another sense, this is consistent with law enforcement decision making goals to resolve
unknown circumstances and attribute causality to specific behaviors.
Prospect Theory
Tversky and Kahneman (1974) developed prospect theory, which holds that
people make choices mostly by how they are framed and that they select alternative
choices by assessing risk where the probabilities are known. The focus of prospect theory
is an estimation of how choices are made under risk. It holds that intuitive judgments
occupy a position between the automatic operations of perception and the deliberate
operations of reasoning. Contained within the prospect theory is the certainty effect
(Kahneman & Tversky, 1979), which is a tendency to under-weigh outcomes that are
merely probable, in comparison to outcomes obtained with certainty. In research on
heuristics, Tversky and Kahneman (1974) described three general-purpose heuristics:
availability, representativeness, and anchoring and adjustment. These heuristics are said
to be simple and efficient as they piggyback on basic computations that the human mind
has evolved to make. Each individual heuristic associates with a set of biases that are
departures from normative rational theory. Kahneman and Tversky (1979) note that
people are limited in their ability to comprehend and evaluate extreme probabilities. As a
result, events that are highly unlikely are either ignored or under-weighed.
In the present study, prospect theory helps to explain perception and reasoning as
decisional factors as a belief in guilt is weighed against other explanations.
Social Functionalist Theory
Tetlock (1992) proposed that psychological explanations are ultimately rely upon
functionalist assumptions regarding how people attempt to achieve goals through
thinking, feeling and acting. Further, Tetlock (1992) noted people engage in an
internalized dialogue whereby they anchor certain values and confer legitimacy upon
social practices (p. 452). In a professionalized setting such as law enforcement, certain
commonalities are enforcing the law. Thus, police serve to protect against encroachments
upon shared social values, which are legislated into laws. In the process of controlling
law breakers, Tetlock (1992) suggested that police must protect from exploitation those
who would violate cultural norms, harming others. There are many assumptions
associated with human behavior so the key is to stick to those assumptions that are
reasonable and based upon sound judgment, an issue that is relevant to police conduct in
sound decision making.
Bounded Rationality
The conceptual framework of bounded rationality is most relevant here because
nowhere is it more descriptive where a person makes a rational judgment bound by
limited information (Snook & Cullen, 2007). Bounded rationality helps to describe how a
person’s belief may be biased through the limited information they possess, or by what
strong beliefs they hold. This psychological framework acknowledges inherent
processing limitations of the human mind. People reason and choose rationally, but only
within the constraints imposed by their limited search and computational abilities
(Findlay & Scott, 2006). It is the idea that in decision-making, rationality of individuals is
limited by the information they have, the cognitive limitations of their minds, and the
finite amount of time they have to make a decision. This is particularly relevant to the
present study, as the investigators will have limited time on which they may deliberate on
the stimulus. In this context, bounded rationality is a dynamic concept that can change
with exposure to information.
In former research the concept of bounded rationality has been applied to criminal
investigative decision making in the form of responses to vignettes of carefully crafted
criminal cases (Ask & Granhag, 2005; Dror & Fraser-Mackenzie, 2009; Rassin, 2010).
This method provides an effective framework from which to produce the desired bias for
experimental purposes, and to present case facts in such a manner as to stimulate a
response from the participants about how they feel about their previous answer in light of
newly acquired information.
Naïve Realism
People typically believe that facts as they see them, are there for all to see and
therefore, others should agree with them. Naïve realism posits that if people do not agree
with another person, they believe those others have either not been exposed to the
relevant facts, or they are blinded by their own interests and ideologies (Pronin, Ross, &
Gilovich, 2004). In the present study this can help to describe how an investigator may
believe their perception is accurate because, in their eyes they possess a unique insight.
While this may be true, the investigator’s belief is not as important as what they may be
able to prove from an evidentiary standpoint.
Bias Blind Spot
The bias blind spot is a phenomenon where people are motivated to view
themselves, their assessments, and their outcomes, in a positive light. A person’s
selfperception (like social perception) causes them to rely on, and attend to, information
such as their own private mental events, where they believe they have more precise
information than the observer about their own emotional states and intentions (Pronin,
Gilovich, & Ross, 2004). We know that people perceive the world in their own way and
they accept that others’ senses are different than theirs, however, they tend to believe that
their senses and perceptions are objective and that it is others whose perceptions and
objectivity are altered. When people’s knowledge of their own intentions is direct, other’s
knowledge of those intentions usually quite inferior and highly subject to error (Nisbett &
Ross, 1980). Many people decide that what is normal is what they would do, when they
would do it, and they cannot or will not accept anything else. In Pronin, et al., (2001)
research participants overwhelmingly reported that they were personally less susceptible
to each of these biases than the average American.
People are often confronted with instances in which others respond to issues and
events differently than we do. We tend to resolve people seeing the world differently than
we see it by declaring that those others, as a consequence of some basic intellectual and
moral defect, are unable to see the things “as they really are” and to react to them “in a
normal way” (Pronin, et al., 2004, p. 783). The implication is that things are in fact as we
see them, and that our ways are in fact normal.
Notably, people expect other reasonable and attentive people to perceive the same
reality they do and cannot fully escape the conviction that the perception of such events is
shared by other reasonable people who have the same information about those events and
issues who will, or at least should, perceive them similarly (Pronin, et al., 2004). Most
people can recall instances in which they have accepted evidence that was at variance
with their preconceptions and that they would have preferred to reject. In fact,
introspection often produces the honest conviction that one acted as one did in spite of,
not because of, one’s private sentiments. Wilson (2002) suggested, more generally, that it
is the very amount of inside information that we possess that gives rise to misplaced
confidence in its authority and authenticity as a guide in assessing the causes of one’s
own behavior. The extent to which the actor behaves differently in the same situation
should cause each to attribute the other’s actions to internal dispositional factors.
Naïve realism and the bias blind spot are two particularly relevant theories in this
study that explain how faulty decision making occurs, and how an individual’s judgment
regarding evidence affected through a person’s personal beliefs, and may be in
contradiction to the actual meaning of evidence. In particular, Wilson (2002) discussed
how inside information (i.e., evidence) can give a false sense of authenticity (i.e., guilt
belief) to the decision maker (i.e., criminal investigator).
Information Processing Theory
Cognition is the focal point of social psychology and has many concepts in
experimental psychology. One early psychologist whose work led ultimately to social
cognition research was Kurt Lewin (1951), who imported gestalt ideas from Germany
that emphasized the influence of one’s social environment as perceived by the individual.
Referred to as the psychological field, this individual perception describes how a person
perceives and understands their own environment based on the total situation, and not
isolated elements, and also being independent of others since what matters is the person’s
own interpretation (Fiske & Taylor 2013). The present study builds upon the emphasis of
information processing theory on the real-world social issue of criminal investigative
decision making. Nickerson (1998) discusses information processing for the confirmation
bias saying that it is sometimes attributed to the tendency of people to gather information
about only one hypothesis at a time, and that the hypothesis is only true (or only false),
but neither probability could be considered simultaneously. So the information processing
theory limits the decision maker to one hypothesis, and assigns only one truth or
falsehood to it. Lord, Ross, and Lepper (1979) found that people rated confirming
information as more convincing that disconfirming.
Intuition
Kahneman (2003; 2011) described human intuition as automatic and effortless and
fast and powerful. Reasoning, on the other hand, is more slow and deliberate and is thus
more controllable and subject to interference. Intuition is an implicit function that humans
learn and is thus difficult to modify or control (Kahneman, 2003; Rossmo, 2009;
Sternberg, 2009), Although intuition operates at a below-consciousness level, it is still
based upon normal sensory input (Douglass, 2008; Rossmo, 2009). Intuition is often
emotionally charged, implicit, and accessed easily through the use of heuristics. Intuition
can help perform a number of tasks quickly and correctly, however, intuition is also often
wrong (Kahneman, 2003). Using intuition is easier than reasoning, and in complex
situations humans search for the most easily accessible solution. Because it is more
thoughtful and deliberate, reasoning can overrule intuition.
Heuristics and Biases
Kahneman and Tversky (1982), inspired by examples of biased, real-world
judgments, further developed on the theory of bounded rationality. Heuristics are “mental
shortcuts” that are informal, intuitive, speculative strategies that most often lead to
effective solutions (Sternberg, 2009). The heuristics and biases approach developed by
Kahneman and Tversky (1984) recognized that intuitive judgments were categorically
different than simple rational choice models (Gilovich & Griffin, 2002). In the process
they described three general-purpose heuristics: availability, representativeness, and
anchoring and adjustment. These heuristics, each associated with a set of biases, are
simple and efficient and have a tendency to integrate with judgments and decisions that
need to be made.
One feature of heuristics is that they can yield quick solutions while drawing on
sophisticated, underlying processes (Gilovich & Griffin, 2002). It is important to note that
biases and heuristics do not always yield optimal or even accurate decisions, due to
excessive complexities. One way to understand this approach is to think of the human
mind as a computer with limited processing. The response is efficient and but not always
accurate. In this context heuristics and biases should be acknowledged as potential
influences on important decisions, and include expectations and preconceptions that can
affect criminal investigative decision making, and ultimately the strength of belief in a
suspect’s guilt.
The information processing and heuristic and bias approach describes heuristic
processing information with very little effort in judging it’s validity (Chaiken, 1980),
relying upon more accessible information or non-content cues, and preferring
confirmatory information (Lord, Ross, & Lepper, (1979) when determining the
conclusion of the message.
Anger Attributional Affect
The attributional effects of anger are expected to support the theory that anger
may lead to stronger attributions of blame. Quigley and Tedeshi (1996) concluded that
once experiencing an anger-type emotion, an investigator might be more inclined to
search for evidence of guilt and believe in a suspect’s culpability. This type of guilt
judgment is likely to inspire heuristic processing that may lead to even more blame
attribution. Further research supports the concept of anger attribution (Maroney, 2006;
Young, Tiedens, Jung, & Tsai, 2011).
Anger also leads to the propensity to seek confirmatory information (Young, et.
al, 2011). In the present study it is theorized that anger may lead to attribution of blame
toward a suspect in an emotional crime such as child sexual abuse. In the present study it
is hypothesized that a criminal investigator may suspend skepticism and arrive at a belief
in guilt before it is justified by thorough investigation of other evidence or
disconfirmatory information. This behavior has the potential to exacerbate confirmation
bias through anger attribution.
Police Organizational Culture
Examining police decision making from the perspective of organizational culture
contributes to the understanding the effect of law enforcement environments (Morreale,
Bond & Dahlin, 2003; Paoline, Myers & Worden, 2000). Organizational culture can be
characterized by suppositions, shared beliefs, opinions, and prototypes that are taken for
granted without questioning and shared by members within an individual agency
(Jermier, Slocum, Fry, & Gaines, 1991). A “culture” can possess both negative and
positive attributes. The U.S. Department of Justice (1997) described police culture as a
“set of norms or beliefs that guide a particular group's behavior, the truths that officers
feel in their bones, the touchstones that govern their attitudes and behavior..." (p. 30).
Thus, culture is deeply ingrained in law enforcement ethos.
In the present study, an understanding of police culture will help to properly
recognize the contribution law enforcement officers can add to the validity of this study.
The responses in this study will include the important aspect of “tacit experience” that
police acquire by virtue of their experience. Pozzali (2006) explains that not all scientific
knowledge can be expressed in explicit form. Much of it is stored in the practices,
abilities, personal insights and expertise of the individual practitioner, none of which can
be easily codified. Tacit knowledge has been considered a type of bias, however, like
riding a bike, it expresses a reality that is part of the application of implicit knowledge
(Polanyi, 1958). Police culture can help to give some context to the unique form of
decision making that occurs in the criminal justice system.
The Blue Wall of Silence: In police culture “The Blue Wall of Silence” is a
colloquial phrase that reflects what is police perceive is a relentless war on crime, with
individual loyalty, solidarity and uniformity among police officers, coupled with the
belief that secrecy against outsiders is necessary to maintain it (Cancino & Enriquez,
2004; Crank, 1998; Goldsmith, 1990; Manning. 1995; Paoline et al, 2000; Kingshott et
al., 2004). There is a belief by many officers that police work cannot be fully understood
by outsiders (Crank, 1998) and people who are not there to observe the tacit and
intangible aspects of interaction with the criminal element cannot fully appreciate the
nature of the work. While the public demands more and more from police, the perception
is that the public does not support them enough (DOJ, 1997). When a rogue officer
commits a crime or when a wrongful conviction occurs it becomes nation news and
media will represent that such instances occur more frequent than they really do,
insinuating that police are less trustworthy. This contributes to an us versus them
mentality among officers (Kingshott, Bailey, & Wolfe, 2004; Cochran & Bromley, 2003;
Crank, 1998; Herbert, 1998; Kingshott, Paoline, Myers, & Worden, 2000) creating the
assumption that only other police officers, and not citizens, are trustworthy and can
understand and support fellow officers. Drummond (1973) explained this contributes to
the police axiom "we have to look out for each other" (p. 25).
Solidarity certainly has negative implications regarding the code of silence,
however, the literature focuses heavily upon negative aspects that include illegal activity,
misconduct, excessive use of force, corruption, and unethical behavior (Goldsmith,
1990; Harrison, 1998; Kingshott, Bailey, & Wolfe, 2004; Paoline, Myers, & Worden,
2000; Terril, Paoline, & Manning, 2003). Trautman (2001) concluded that the code of
silence is a source of police deviance that serves as a protective shield for illegal
behavior. Harrison (1998) stated that it is more influential in police than other
professions. Crank (2008) argued that officers will “purify” their abuse of suspects and
illegal behavior through a “police morality” they construct, that justifies what they do on
their own turf (p. 155).
Not all of the research agrees. Solidarity among officers can also have a positive
influence on the police environment (Paoline, Myers, & Worden, 2000). As Harrison
(1998) noted, solidarity can provide a source of pride and partnership that can cultivate
camaraderie and esprit de corps in an agency Likewise, Sever (2008) concluded that
solidarity and can serve to unite against hostility in unpredictable environments. With
such disparate meanings it is uncertain what the rationale is for police silence. What is
certain is that getting input from police in the understanding of psychological
phenomenon is crucial to making progress on issues of that contribute to positive social
change.
It is important to endeavor to work with law enforcement in the attempt to reach
realistic research findings that can benefit agencies in their professional development, and
to better serve their communities. The present research proposes to include police officers
in studying law enforcement decision making rather than relying on convenience
sampling. It is not difficult to understand the cynical nature of police when research is
critical of them, and no attempt is made to examine the nature of their actions. This
misconception prevents collaborative scientific research rather than drawing interested
fields of inquiry together. For this reason it is beneficial to obtain the cooperation of
police as research participants in a matter of such important public interest.
Literature Review
The seminal work by Tversky and Kahneman (1974) and Nickerson (1998)
enlightened the study of social cognition and contemporary research and has been cited in
thousands subsequent research studies. Work in the domain of heuristic and bias research
includes that of Daniel Kahneman and Amos Tversky who published a series of papers
about decision making under uncertainty and risk (Kahneman, 1973; Kahneman &
Tversky, 1972, 1973, 1979, 1986, 1996, 2000; Kahneman, Slovik, & Tversky, 1971,
1972, 1982; Tversky & Kahneman, 1971, 1973, 1974, 1981, 1983, 1986, 2002). This
field-changing work outlined three commonly employed heuristics as good examples of
how people reduce complex tasks into more simple, less effortful operations: availability,
representative, and anchoring and adjustment (Tversky & Kahneman, 1974). This
program of research has become known as the heuristics and biases program (Snook &
Cullen, 2009), explaining how heuristic processes can often sometimes lead to errors and
biases and that people assess and reason about a situation in a way that is very sensitive to
prior information.
Heuristics-oriented decision making refers to the use of simple mental strategies
to deal with dubiety and an uncertain world (Fiske & Taylor, 2013; Kerstholt &
Eikelboom, 2007). One criticism of this research however, is that they imply step-by-step
search strategies (Dhami, 2010). The idea that they are bad stem from research that
strives for optimality (Kahneman, Slovic, & Tversky, 1982; Nisbett & Ross, 1980).
Specifically, this research holds that heuristics ignore certain information that might be
relevant, and do examine information that other idealistic models do (Snook & Cullen,
2009).
Another seminal study included a meta-analysis by Nickerson (1998). In this
study on confirmation bias the question of spontaneous versus deliberated case-building
is addressed. Nickerson (1998) describes deliberate case-building as an attorney’s job,
whereas spontaneous case-building is of the less conscious, more biased sort. This study
goes on to describe motivated vs. unmotivated forms of confirmation bias.
O’Brien (2009) conducted a study of some relevance to this research issue by
using a common methodology; participants were exposed to an element of bias and then
questioned about their feelings of belief in a particular subject’s guilt. Ask, et al., (2008)
have studied the concept of confirmation bias in the context of evidence, and the elasticity
(the latitude for interpretation) of evidence in changing circumstances. Rassin (2010;
2012) conducted research of a similar vein, evaluating the susceptibility of confirmation
bias, and also biased evaluations of criminal evidence. O’Brien (2009) and Ask, et al.,
(2008) and Rassin (2010, 2012) all generalized their findings to the field of criminal
investigation, theorizing that the data from their research participants yielded valid
responses.
The methodologies of these studies included quasi-experimental designs. O’Brien
(2009) divided participants into two conditions (hypothesis, no hypothesis) x 2 groups
(half-case file, full-case file) and utilized a split-half reliability method. Ask, et al., (2008)
conducted a 3 x 2 factorial design with three evidence types (DNA, photos, witness) and
two evidence outcomes (consistent, inconsistent). Rassin (2010) also conducted a
factorial design with two evidence types (severe, aggravated) and two strengths of
evidence (strong, weak). In each study, participants received evidentiary propositions
with certain facts they were to assume were true. The participants also completed short
questionnaires about their impression of guilt and strength of evidence. In addition to the
use of Likert scale questionnaires, this study proposes to conduct hierarchical multiple
regression analysis to examine whether the characteristics and variables such as
education, age, time in grade, gender, or training serve to moderate the effects of
confirmation bias.
Among the most current research that informs this study includes. Busey and Dror
(2011) investigated forensic decision making from an information processing perspective,
and West, Meserve, and Stanovich (2012) explained how the bias blind spot interferes
with optimal decision making in an investigative context. Girgenzer (2010) explained
how moral satisficing could lead to decision making. Kebbell, Muller, and Martin (2010)
investigated decisional bias in dealing with uncertainties in serious crime, and Keel
(2008) identified some “best practices” for investigations.
Research into law enforcement decision making includes an emphasis on
organizational culture (Goldsmith, 1990; Harrison, 1998; Kingshott, Bailey, & Wolfe,
2004; Paoline, Myers, & Worden, 2000; Terril, Paoline, & Manning, 2000: Trautman,
2001). Criminal investigative decision making is influenced by shared beliefs and
opinions between members of an agency (Jermier, Slocum, Fry, & Gaines, 1991). Within
law enforcement agencies, the more experienced officers often shape the beliefs and
opinions of newer officers, creating a more uniform subculture, and perpetuating the
cycle (Sever, 2008). Part of this subculture includes uniformity and solidarity in a war on
crime (Bailey & Wolfe, 2004; Paoline, Meyers, & Worden, 2000), creating a morality
where questionable decisions are “purified” (Sever, 2008), thereby justifying “bending”
the rules in order to be successful against criminals (Cochran & Bromley, 2003). This is a
more conscious, intentional form of bias than what is being considered in the proposed
study, however, it helps to explain how strong the influence of organizational culture.
Research by Fahsing and Ask (2013) studied affective reactions in an
investigative context, and described how sexual offenses violate shared human values and
influence investigative decision making. This study is consistent with the theory that the
types of evidence may lead to different levels of confirmation bias. Quigley and Tedeschi
(1996) showed that anger can transcend a situation and increase blame attribution to
subsequent events. Dror (2008) describes how mental states play active roles in decision
making in whether and how we acquire, process, and interpret information. Thus, even
before information is collected, people will usually have a preference (Busey & Dror,
2011). Capesteny and Harris (2014) examined how more disgusting crimes could
influence reason in legal decision making, including affective responses. Offering more
physiological substance, Greene, Sommerville, Nystrom, Darley, & Cohen (2001)
showed that in participants who considered personal moral dilemmas involving direct
contact with a victim, the brain demonstrated increased emotional activation.
Feigenson & Park (2006) explain that in a forensic context, emotion includes
feelings, cognitions, and actions (or inclinations to act). Kruglanski and Gigerenzer
(2011) demonstrated how it emotionally significant crimes lead to satisficing in legal
decisional judgment. That is, finding a way to conclude guilt with the existing evidence.
Additional research offers support for the research hypothesis that emotion can result in
forming a guilt judgment (Bornstein & Wiener, 2006; Maroney, 2006; Weiner, Bornstein,
& Voss, 2006). Finally, once a guilt judgment is made it can be cognitively adopted as
factual (Dror & Fraser-Mackenzie, 2005), and once adopted as factual the judgment is not
likely to be amended.
According to Maroney (2006), the literature on emotion in legal decision making
reached a “high-water mark” with a series of essays from scholars across many
disciplines (Bandes, 1999), however, psychology was excluded. Though advancing the
dialogue, Bandes (1999) did little to explain the cognitive content of emotion in legal
decision making. Nevertheless, this prompted further study on the subject of emotion in
law and there is substantial support for the concept of anger attribution. More recently,
Stevenson, Malik, Totton, and Reeves, (2014) demonstrated how extreme emotion can
cause jurors to dehumanize a super-predator defendant who perpetrates what they
perceive as disgusting crimes.
Research by Ask and Granhag (2007) investigated similar outcomes of interest
using a methodology similar to the present study. Schrackmann and Oswald (2014)
discussed the preference for confirmatory information in an investigative setting when
testing attitudes or beliefs. This research allowed participants to revise their level of
suspicion toward a suspect after reviewing evidence. Additional research supporting the
hypothesis of emotion influencing investigative decision making is found in
Bodenhausen (1992), Ditto, et al., 1998; 2006), Kosnik (2007; 2008), Klayman and Ha
(1987) and Wilson (1998).
The presence of confirmation bias in criminal investigation is supported in
multiple studies (Ask & Granhag, 2005; Ask, et al., 2008; Busey & Dror, 2011, Dror &
MacKenzie, 2009; Eerland & Rassin, 2010; O’Brien, 2009; Rassin, 2008; Rossmo, 2009;
Snook & Cullen, 2009; Spano, 2005). These studies are relevant to the research
hypotheses that confirmation bias can be present in criminal investigations, but there may
be individual characteristics that moderate the phenomenon.
Among the studies of confirmation bias and criminal investigations, those that are
most central to the current study are O’Brien (2009) and Weiner, Bornstein, and Voss
(2006). Ask and Granhag (2005) and Rassin (2008; 2010) are often cited in the literature
as key studies, however, they are not included in the same category here based upon the
non-representative choice of research participants as well as the broad differences in
Swedish and Dutch criminal justice systems compared to the U.S. O’Brien (2009) also
used convenience sampling, and with 18-year old college freshman might be argued that
they are even less representative, however, O’Brien (2009) is the only research identified
in the U.S. on the subject close to what the present study proposes.
Of the studies that bear directly upon the issue of confirmation bias in criminal
investigations, O’Brien (2009) considered that if experienced criminal investigators
demonstrated confirmation bias, the natural follow-up questions are what exacerbate it,
and what might mitigate it? Based upon suggestions that deploying countermeasures
might force an investigator to consider alternative hypotheses, O’Brien (2009) suggests
that thinking of reasons why their hypotheses might be wrong, or that another suspect
might be viable, might diminish an investigator’s bias. The research is well
conceptualized except it branches off here and focuses on showing how to “reduce” the
tendency toward theory favoritism.
The greatest weakness of this research is in the use of lay personnel as
participants. Even O’Brien cautions that the findings of the study are “limited by the lay
participants”. This weakness is understated, however, as the average age of the
participants was 18-years old and they all received course credit in exchange for their
participation. Additionally, the study went to great lengths to replicate a complicated
homicide case with complex evidence; a type of case that would require years of relevant
experience to acquire the nuances and subtleties in understanding case complexities,
along with the importance of disconfirmatory evidence and the legal requirement of
reporting exculpatory evidence. An inexperienced layperson runs the risk of categorizing
something immediately if it makes reasonable sense to them, whereas criminal evidence
can quite often only makes sense in retrospect, if at all.
Rassin, et al. (2010) and Ask, et al. (2008) are studies of direct relevance that
arrived at different conclusions than O’Brien (2009). Rassin, et al. (2010) found that it is
difficult for an investigator to remain open to alternative scenarios, even if they clearly
exist. Ask, et al., (2008) investigated the phenomenon by examining different kinds of
evidence (witness, photo, DNA) and whether inconsistent evidence might be considered
less reliable than consistent evidence. This study found that the reliability was more
pronounced when the evidence consisted of less interpretation such as DNA, and the
evidence did not challenge their suspicions. When the interpretation of the evidence was
greater, such as witness statements of photographic evidence, the probability of
contextual influences was also found to be greater. This research provides insight into the
evaluation of certain forms of evidence, but does not delve into investigative decision
making or any of the different forms of bias.
Ask and Granhag (2007) examined the differential impact of anger and sadness on
criminal investigative judgments. Researchers there considered the concept of strong
affective reactions to child sexual abuse, a crime that provokes strong emotions. The
researchers focused on judgments made at the early stage of a criminal justice process
(criminal investigation), and considered the resultant consequences of anger and sadness,
hypothesizing that anger may transcend the situation and increase blame attributions
related to the crime. Ask and Granhag (2007) noted that previous research on anger and
judgment show that participants make more stereotypical judgments (Bodenhausen, et al.,
1994) and may rely on fewer diagnostic cues (Lerner, Goldberg, & Tetlock, 1998).
Tiedens (2001) noted that in such cases participants pay closer attention to superficial
clues and make greater use of chronically accessible information (Tiedens & Linton,
2001). Ask and Granhag (2007) concluded that anger is likely to make criminal
investigators rely on superficial processing, and that they may base their judgments more
on preexisting expectations and beliefs, than on deeper processing.
Fahsing and Ask (2013) conducted research on investigator’s decision making,
hypothesizing that time pressures, emotional involvement, and expediency-promoting
occupational norms (time) are the primary obstacles of optimal decision making.
Kruglanski and Webster (1996) defined the ambiguity of an unsolved crime as a strong
need for cognitive closure, motivating the investigator to come to a conclusion rapidly. In
another study that informs a hypothesis of the proposed study that emotion can affect
decision making, Wiener, Bornstein, and Voss (2006) found that emotions experienced at
the time of a legal judgment can help to determine the decisions and behavior. This study
concludes that a decision need not arise from the stimuli that occur before the decision,
but that a decision may be influenced by internal or external factors that are incidental to
the task.
One limitation of many of the aforementioned studies was a lack of internal
validity. Virtually all of the research on criminal investigative confirmation bias has been
conducted with participants with very little life experience, and no law enforcement
experience, upon which the relevant perspectives rely. Ask, et al. (2008) explained that
criminal investigations include complex contextual influences and that background
knowledge contributes significantly to decision making and judgment. The use of student
participants obscures the actual dilemma confronting the perceiver who encounters
confirmation bias stimuli. In this regard, 18-year old college students are not
representative of experienced law enforcement officers making complicated decisions in
dynamic situations. This use of lay participants has been addressed as a limitation (Ask,
et al., 2008; O’Brien, 2009; Sears, 1986) with recommendations that experienced
investigators be utilized in the future for more accurate interpretation of the phenomenon
(Rassin, et al., 2010).
Another weakness of these studies is that quasi-experimental designs preclude
firm causal inferences (Frankfort-Nachmias & Nachmias, 2008). The research by Ask, et
al. (2008) and Rassin (2010) were conducted in countries with different socio-political
cultures whose legal systems have varying levels of proof that are incompatible with the
US criminal justice system. In the U.S. judgments of guilt vary according to the required
level of proof and are required for a different rational as well. For example, in the U.S.
probable cause is required to conduct a search or effect an arrest. Another notable
difference in legal systems is that Sweden allows hearsay evidence whereas it is strictly
limited in US courts. Also, Swedish Judges may consider alternative evidence not
presented in evidence. Further, Swedish courts are not bound by prior decisions, limiting
consistency in the application of the law. Law enforcement in Sweden, like government,
is centralized, and run by the Ministry of Justice. In the U.S. law enforcement is more
local with 58% of all agencies smaller than 50 officers or less (Reeves, 2010). Civilian
oversight for local agencies exists within the communities that they serve.
Rassin’s (2010) research with Swedish police officers, lawyers and judges,
compared emotion (anger and sadness) with statement consistency (consistent and
inconsistent). This research concluded that sad investigators engaged more thoroughly in
processing case material than the investigators who were angry, and that sad participants
were more likely to believe the suspect was guilty. In research by Rassin, Eerland, and
Kuijpers (2010) law students were used as participants and were found to be more likely
to prefer evidence of an incriminating or exonerating nature, depending upon their
preliminary beliefs in innocence or guilt. The conclusion of this research stressed the
importance of delaying conclusions of guilt or innocence until all of the information is
obtained. The findings of this research are in line with the research of others (Ask &
Granhag, 2007, Nickerson, 1998, O’Brien, 2009) that people have a tendency to look for
evidence that confirms their beliefs.
The effect of anger and emotion on criminal investigative judgments is highly
relevant as such judgments may influence the investigator’s attributional judgments
towards personal and situational causes (Ask, et al., 2008; Dror, 2005). This may come in
the form of affecting the perception of the causes a perpetrator’s behavior, or the
consequences for a victim. In the present study, the effect of anger is anticipated to come
in the form of a swifter guilt judgment. It is also anticipated that empathy for a minor
victim may also contribute to a swifter guilt judgment.
In studying confirmation bias Ask, et al., (2008) considered the concept of
asymmetrical skepticism and utilized as an IV the type of evidence (physical evidence,
testimonial evidence, DNA) and compared it to a level of belief in guilt as a DV. Rassin
(2010) did the same. In the present study, similar variables were chosen, as they are both
relevant representative to the issue under investigation. No study into confirmation bias in
criminal investigation would be complete without bias or suspicion toward the suspect.
This study utilized a level of bias as an IV, and a measure of guilt judgment toward a
suspect using Likert scale responses as a DV.
A key component missing in the existing research regarding confirmation bias in
criminal investigative decision making is the applied nature of the decision maker. The
applied nature of the criminal investigator incorporates tacit variables that are uniquely
and intrinsically present the law enforcement officer, based upon their training and
experience. Without an applied component, research distills the information such that the
layperson must consider the findings in a vacuum and consider how they might be, rather
than with an applied perspective that more accurately demonstrates how they really are
(Pozzali, 2006). This study sought to provide that applied component.
Summary
In this chapter I reviewed the literature on confirmation bias and forensic decision
making. This review included relevant research discussing confirmation bias in criminal
investigation as well as the effect of emotion on judgment and decision making. I have
also summarized relevant theories and concepts used by researchers including rational
choice, expected utility, prospect theory, and relevant framework from bounded
rationality, heuristics and biases, intuition, information processing, anger attribution,
social functionalist theory, and police organizational culture.
Together, the studies outlined offer preliminary evidence that confirmation bias
does exist in criminal investigative decision making, but there are some institutional
safeguards and individual characteristics that can moderate such bias. Further, these
studies have shown that when bias takes place in criminal investigative decision making,
it can persist despite the presence of disconfirmatory evidence. These are problematic
areas of concern in a forensic setting such as law enforcement.
Confirmation bias is a ubiquitous phenomenon in human nature (Nickerson, 2008)
that causes people to selectively seek information that is consistent with prior beliefs,
expectations, or desired outcomes, to disregard, deny, devalue or dismiss information that
is inconsistent or disconfirmatory, or to interpret ambiguous information in a manner that
confirms (or disconfirms) a person’s perceived beliefs or expectations (Dror &
Mackenzie, 2005; Lilienfeld, 2010; Wiener, Bornstein, & Voss, 2006). The literature on
confirmation bias in criminal investigative decision making has demonstrated the
presence of confirmation bias (Ask, et al., 2008; O’Brien, 2009; Rassin, 2008; 2010).
Rassin (2010) notes that an investigator may create search strategies in such a way that
confirmation is given more of an opportunity than is falsification.
It is a basic assumption of public conscience that the police operate fairly,
impartially and objectively, making decisions only after they have collected the entire
body of evidence. Preexisting attitudes, expectations, biases, preconceived notions,
stereotypes and schema are variables that may affect criminal investigative decision
making. Thus, in the process of investigating a case, and investigator may seek only that
evidence which inculpates the defendant, building a case rather than following the clues
and facts. In doing so, one tends to believe theory before they have the facts, thus the
theory drives the fact-finding, and case building ensues.
One notable gap in the literature related to this study is that there has been no
practical analysis of the phenomenon of confirmation bias in criminal investigations with
law enforcement personnel within the United States criminal justice system. This study
will expand on this gap within the literature, studying the effect of confirmation bias on
criminal investigative decision making in an applied manner that affects the whole of
society. The use of law enforcement personnel as research participants will allow for a
more direct comparison of data with the research problem.
O’Brien (2009) conducted research on this topic using college freshman with an
average age of 18-years old. One limitation of this population is that the obsequious
entertainment industry touches young, impressionable people now more than ever with
the proliferation of media and electronics over the last decade. One of the most popular
entertainment subjects is crime drama, which displays inaccurate and unrealistic
depictions of criminal investigations. In 2010, a quarter of the most watched shows in any
given week involve crime, investigation, and forensic science: CSI, NCIS, NCIS: Los
Angeles, and Criminal Minds (Gabel, 2011). This information tends to taint the belief
system and expectations of most people, particularly inexperienced and easily influenced
teens. The present study proposes to utilize law enforcement officers who are assigned to
patrol and investigative assignments Tapping into the tacit knowledge associated with
this experience is expected to provide somewhat more subdued, but more realistic
expectations and beliefs.
The literature reviewed for this research provides insight into the phenomenon of
confirmation bias in a general forensic sense. The existing literature emphasizes the
pervasive nature of confirmation bias (Nickerson, 1998; Ditto & Lopez, 2002; Ditto, et
al., 1998). The relevant theories and psychological frameworks contribute to
understanding the nature of the unique influences and pressures on decisions that must be
made by police officers in times of stress and uncertainty (Dror & Fraser-Mackenzie,
2009; Dror & Bucht, 2011; Epstein, 2004). The research questions built upon this
framework to examine the nature of the influence of emotion and bias on decision making
in an applied, criminal context, by comparing the influence of emotion on a confidence
assessment of guilt judgment. A comparison of duty assignments expounded on whether
criminal investigators were less influenced than others, as the nature of their position
entails. The gaps in the articles identified in the research have laid the foundation for the
framework of this proposed study into confirmation bias, which examined the forensic
aspect of the phenomenon further into the judgment of individual criminal investigators
in decision making.
Despite the problematic areas of human decision making and cognition errors, it
can be argued that logical reasoning can be used to override such psychological biases
(Snook & Cullen, 2005). In light of that, it is worthwhile to expand the field of
knowledge in criminal investigative decision making to become more aware of these
issues, and develop strategies to lessen or reduce bias through accountability protocol,
training, and agency policies, to add to what we know about the phenomenon, and
discover more about what we don’t.
Chapter 3 will focus on the study’s methods and design, providing an explanation
of the design, justification, and measurements utilized in the study, as well as an
extensive analyses of the data. This will include an explanation of the rationale for the
study’s design, a description of the population and sample to be studied, the sampling
procedure, data collection, instrumentation and data analysis procedures. Chapter 3
concludes with a summary of the steps to be taken for the ethical protection of the
research participants.
Chapter 3: Research Method
Introduction
The purpose of this study was to examine the association between confirmation
bias and criminal investigative decision making and to explore the influence of bias in a
practical application. Independent variables (IV) under consideration in this study include
the type of duty assignment (i.e., recruit, patrol and investigations), the type of crime
(child sexual abuse, abuse of adult), and the order of evidence presentation (sequential,
simultaneous, reverse sequential). The dependent variable (DV) in this study included a
degree of confirmation bias, represented by a confidence assessment of a suspect’s guilt
or innocence, in light of specific items of information or evidence.
This chapter describes the methodology employed in this study of the effect of
confirmation bias on criminal investigative decision making, and the researcher’s
justification for the choice of methodological approach. In particular, this chapter
describes the method for studying whether confirmation bias differs by duty assignment,
and whether emotion influences the belief in a suspect’s guilt in a criminal investigative
setting. This chapter also explains the study’s rationale and research design, the
population and sample, instrumentation and materials, procedures for data collection and
analysis, and steps taken for ethical protection of participants.
Research Questions and Hypotheses
A quantitative research design was used for this study, which was organized
around three research questions and associated hypotheses:
RQ 1: Does confirmation bias differ among various assignments (recruit, patrol,
investigators) within law enforcement?
H0 1: Confirmation bias does not differ among assignments.
H11: Confirmation bias does differ among assignments.
RQ 2: Can emotionally charged crimes increase belief in a suspect’s guilt?
H0 2: Emotionally charged crimes do not increase belief in a suspect’s
guilt.
H1 2: Emotionally charged crimes can increase belief in a suspect’s guilt.
RQ 3: Does the order of evidence presentation influence belief in a suspect’s
guilt?
H0 3: Evidence presentation order does not influence guilt belief.
H13: Evidence presentation order does influence guilt belief.
Research Design
The purpose of this study was to evaluate the effect of confirmation bias on
investigative decision making. Further, utilizing a law enforcement body of research
participants provided greater construct validity and allow for further research to build
upon this study. To that end, the research design that was selected was a quantitative,
quasi-experimental between-subjects factorial analysis, believing that using police
officers as research participants in a law enforcement task would result in greater validity
provided by greater statistical power. The power of statistical test refers to the probability
that it will lead to the rejection of the null hypothesis. Any failure to conduct analyses of
statistical power can increase the risk of a false rejection of the null hypothesis (Type I
error) or false retention the null hypothesis (Type II error). The power of the test was
determined based on the test instrument in question, along with information regarding the
sample population. Power, therefore, was dependent on the significance criterion,
research sample, and the population effect size (Cohen, 1992).
Setting, Population, and Sample
The setting for this research study was a state-government-operated criminal
justice training center located in a Midwestern state. Sworn officers from around the state
who gather to attend annual in-service training was invited to participate in the study.
Officers attend in-service training according to their duty assignment (patrol and criminal
investigations). Using a state criminal justice training department will provide access to
experienced criminal investigators who are assembled to attend advanced training. Basic
training recruits at the academy will also be invited to participate.
The target sample size for this study was derived from a power analysis of
GPower 3 (Faul, Erdfelter, Buchner & Lang, 2009). This study anticipates an alpha level
of 0.05 and a medium effect size and power level. More participants than are suggested
was invited to participate to account for possible attrition during the data collection.
A two-tailed, medium effect of .8 and an alpha level of 0.05 was utilized. An
appropriate sample size was utilized with these figures and using GPower 3 software
(Faul, et al., 2009). A medium effect size was so that inferences that can be made about
the relationship among the variables in this study. Further, this study will utilize a
splithalf method to construct two scales. The scores from the questionnaires was
correlated as estimates of reliability, with a correlational coefficient between 0.00 and
1.00. The product (coefficient) will express how much each response (variable) is related.
The resultant predicted power was evident.
Design
This research was considering forensic, social cognition that include biases,
heuristics, prejudices, preconceptions, attitudes, stereotypes, to name a few, and how it
influences criminal investigative decision making. Influence indicates the context of
thought processes or acts of thinking. The theory in this research is that there are errors in
thought processes that are unique to criminal investigation, such as preconceptions or
stereotypes against sex offenders. These thought processes could be harmful when they
give the impression of being right but are actually not.
Griffin and Tversky (1982) demonstrated that the balance of arguments for and
against competing hypotheses determines people’s confidence, with insufficient regard
for the weight of the evidence. Griffin and Tversky (1982) further noted that the
formation of belief and the weighing of evidence are basic elements of human thought
and observed how expert prediction is “often wrong, but rarely in doubt” (p. 230). The
impression of being right when one is wrong can be harmful to criminal investigative
decision making. For the proposed research the data analysis instrument will include a
self-reporting Likert scale questionnaire that will yield responses to fixed-alternative
expressions ranging from 0 to 10, with answers ranging from not guilty at all to
absolutely guilty. The statements in the questionnaire are in Appendix H.
Once the responses are calculated, it is expected they will produce ordered
responses of discriminative power (DP). The items with the highest amount of DP are
considered the items that the best discriminate among the individuals expressing different
attitudes toward the attitude being measured (Frankfort-Nachmias & Nachmias, 2009). In
this way, we obtain the most reliable data possible, to reflect the attitude in question.
After each item of evidence, participants will record a confidence assessment of
their belief of guilt/innocence on a Likert scale. The guilt belief scores was measured as
the difference between the number of items (mean evaluation) and tested against zero for
significance. This study was a quasi-experimental, 2 (type of crime: adult sex assault,
child sex assault) x 3 (assignment: recruit, patrol, investigator) x 3 (evidence presentation:
sequential, simultaneous, reverse sequential), between-subjects, factorial design. The IV’s
included the duty assignment and evidence presentation order, while the
DV included a confidence assessment in the form of level of guilt belief.
Another aspect of this study examined whether emotionally charged evidence
could influence belief in a suspect’s guilt. Participants were advised they were
investigating circumstances surrounding a crime and were receiving several items of
evidence. A list of the evidence is found in Appendix A. Both groups received the same
vignette of a crime involving a child sexual assault. One group received each item of
evidence sequentially, recording a confidence assessment of guilt after each item. The
other group received all of the evidence simultaneously, recording one confidence
assessment of guilt. The predictor variables included the items of evidence and the
criterion variables included the assessments of belief in guilt.
To evaluate the practical nature of confirmation bias in law enforcement in
criminal investigations it is necessary to consider the tacit knowledge of criminal
investigators and law enforcement officers. Pozzali (2006) reported that the application of
tacit knowledge such as the subtleties and nuances of a criminal investigation are often
biases themselves. Consequently, the order of evidence included a reverse sequence
presentation order, to explore the nature of confirmation bias when exculpatory evidence
precedes the inculpatory evidence. No reverse simultaneous order was utilized because
there is effectively no order of presentation when all of the evidence is visible at one time.
The effect was expected to be the same as the existing simultaneous presentation order.
The demographic information listed in Appendix I, was collected at the beginning
of the study and provides basic information related to individual, organizational, and
professional characteristics.
Instrumentation and Materials
Participants were asked to complete questionnaires electronically and record their
answers to the vignettes they read. The vignettes include similar sexual assault scenarios
and come from actual criminal cases investigated by the researcher, supporting the
validity of the as the material, since the scenarios have actually been encountered in a
criminal investigation setting. The following case vignettes conveyed the stimulus
material.
Vignette Experiment 1
Participants were advised they were investigating circumstances surrounding a
reported sexual assault of a minor, along with 10 items of evidence. One group received
all 10 items of evidence simultaneously for their review, and asked to render a single
opinion after examining the evidence. Another group was provided the evidence
sequentially, and asked to render an opinion after each item. A third group received the
evidence in reverse-sequential order and asked to render an opinion after each item. All of
the groups were asked to provide a guilt estimate at the end of the survey in the form of a
guilty or not guilty selection. The vignettes read:
John Doe has been down on his luck recently and he has been staying with his
brother and sister-in-law, Ron and Jo Doe, and their 14-year old daughter Jane. John has
been living with the family for the past several months as has had trouble finding a job,
due in part to the fact that he is a convicted felon. Jane reports that John raped her, and
has been doing so for a while. You have been assigned to investigate the case. There are
10 items of evidence. Choose the number that corresponds to your level of suspicion
about the report from 0 (not guilty at all) to 10 (definitely guilty).
Description of Evidence – Sequential and Simultaneous
1. Initial report – You are dispatched to the hospital to meet Ron and Jo
Doe.
You learn that their 14-year old daughter, Jane Doe, reported that her paternal uncle, John
Doe, had raped her. John Doe has been living in the home for 6 months now. Jane’s
mother states that Jane is not sexually active and would not willingly have sex with
anyone, least of all her uncle.
2. SANE nurse report of examination - The SANE nurse notes in her
report that the victim claims to be sexually active with her boyfriend. The report indicates
the victim stated her uncle vaginally raped her and had been doing so for several months
now. The report also notes that the victim reported to the SANE nurse that she is
concerned about possibly being pregnant. The victim was given the option of an
abortifacient (Day after) pill, which she took. No signs of force or injury were noted, but
signs of recent sexual activity were present.
3. Crime scene - Utilizing an alternate light source, body fluid was identified
on the sheets in Jane’s bedroom. Preliminary lab results indicate the substance may be
semen. The sample was submitted to the lab for analysis.
4. Victim interview - Jane Doe states that her Uncle, John Doe, had been
flirting with her and that their rooms are so close in the house that she is uncomfortable
with him being around, particularly since he is a sex offender. A criminal history review
confirms that John Doe is a convicted sex offender, however a check of local records
reveals he has not registered on the local Sex Offender Registry.
5. Suspect interview – John Doe denies the allegation. John Doe states that
Jane is vindictive and resents him being in the home. He states the proximity of his room
to Jane’s restricts what she can get away with and she dislikes him being there. John Doe
states he has caught Jane and her boyfriend having sex in the house on several occasions.
On one occasion when he caught the boyfriend sneaking into her room he confronted
them. At that time Jane threatened that she would tell her parents that he forced himself
on her. John Doe states that his felony conviction is related to having sex with his 16-year
old girlfriend when he was 18-years old. John Doe relates that he is aware he is obligated
to register on the local sex offender registry but has not yet done so because was only
staying temporarily with his brother and had not yet established a residence of his own.
6. Boyfriend interview – The boyfriend states that he and Jane have sex
almost daily and they had sex on the morning of the reported incident. He further states
that he did not use a condom. The boyfriend states that he is not aware if Jane is on birth
control or not, but that she was worried that she might be pregnant.
7. School counselor - Jane had come to her about a week before the reported
incident and asked for information about an abortion. Jane was concerned she might be
pregnant and asked how much an abortion would cost? The school counselor told Jane
that in cases of rape, an abortifacient is free.
8. Follow-up investigation – In a follow-up interview with John Doe. He
reports that he had a vasectomy years ago, thus he no longer produces semen in his
ejaculate.
Medical records provided confirmation.
An interview with a neighbor reveals that they have observed someone coming in
and out of an upstairs window at the Doe’s house, using the roof for access. No
identification of the individual was possible.
9. Follow-up investigation - In a follow-up interview with Jane, she stated
she just wanted the whole thing to “go away” and that she no longer wanted to pursue
charges against her Uncle.
10. Lab report - DNA from the sheet was matched to Jane and included
semen from an unidentified male subject.
Recording Responses – Sequential and Simultaneous
Participants in the sequential group will record their confidence assessment on a
Likert scale after each item of evidence is presented. Participants in the simultaneous
group will record a single confidence assessment on a Likert scale after all 10 items of
evidence are presented. The confidence assessment were measured with the following
11point Likert scale:
0 1 2 3 4 5 6 7 8 9 10
Some of the evidence is framed in such a way that it is directed at obtaining
further evidence incriminating the suspect, implying or confirming the suspect’s guilt.
Other items of evidence are framed in such a way that exonerates the suspect by reducing
the strength of the existing incriminating evidence or by suggesting evidence for an
alternative scenario. At the conclusion of the vignette, the participants will also be asked
to provide an overall estimate of guilt in the form of yes or no selection.
Description of Evidence – Reverse Sequential and Simultaneous
The reverse-sequential and simultaneous groups will receive the same items of
evidence in a different sequence. Some of the verbiage is reworded for logic and
coherence.
1. Initial report – You are dispatched to the hospital to meet Ron and Jo
Doe. You learn that their 14-year old daughter Jane Doe, said that her Uncle John had
raped her. John Doe has been living in the home for 6 months now. Jane’s mother states
Jane is not sexually active and would not willingly have sex with anyone, least of all her
Uncle.
2. SANE nurse report of examination - The SANE nurse notes in her
report that the victim claims to be sexually active with her boyfriend. The report indicates
the victim stated her uncle vaginally raped her and had been doing so for several months
now. The report also notes that the victim reported to the SANE nurse that she is
concerned about possibly being pregnant. The victim was given the option of an
abortifacient (day after) pill, which she took. No signs of force or injury were noted, but
signs of recent sexual activity were present.
3. Crime scene. Utilizing an alternate light source, body fluid was identified
on the sheets in Jane’s bedroom. Preliminary lab results indicate the substance may be
semen. The sample was submitted to the lab for results.
4. Lab report - DNA from the sheet was matched to Jane and included
semen from an unidentified male subject.
5. Follow-up investigation – In a follow-up interview with John Doe. He
reports that he had a vasectomy years ago, thus he no longer produces semen in his
ejaculate.
Medical records provided confirmation.
An interview with a neighbor reveals that they have observed someone coming in
and out of an upstairs window at the Doe’s house, using the roof for access. No
identification of the individual was possible.
6. School counselor - Jane had come to her and asked for information about
an abortion. Jane was concerned she might be pregnant and asked how much an abortion
would cost? The school counselor told Jane that in cases of rape, an abortifacient is free.
7. Boyfriend interview – The boyfriend states that he and Jane have sex
almost daily. Had sex with Jane on the morning of the reported incident. He further states
that he did not use a condom. The boyfriend states that he is not aware if Jane is on birth
control or not, but that she was worried that she might be pregnant.
8. Suspect interview – John Doe denies the allegation. John Doe states that
Jane is vindictive and resents him being in the home. He states the proximity of his room
to Jane’s restricts what she can get away with and she dislikes him being there. John Doe
states he has caught Jane and her boyfriend having sex in the house on several occasions.
On one occasion when he caught the boyfriend sneaking into her room he confronted
them. At that time Jane threatened that she would tell her parents that he forced himself
on her. John Doe states that he was convicted of having sex with his girlfriend, who was
17-years old at the time.
9. Victim interview - Jane Doe states that her Uncle, John Doe, had been
flirting with her and that their rooms are so close in the house that she is uncomfortable
with him being around, particularly since he is a sex offender. A check of local records
confirm that John Doe is on the Sex Offender Registry.
10. Follow-up investigation - In a follow-up interview with Jane, she stated
she just wanted the whole thing to “go away” and that she no longer wanted to pursue
charges against her Uncle.
Recording Responses – Reverse Sequential and Simultaneous
Participants in the reverse-sequential group will record their confidence
assessment on a Likert scale after each item of evidence is presented. Participants in the
simultaneous group will record a single confidence assessment on a Likert scale after all
10 items of evidence are presented. The confidence assessments were measured with the
following 11-point Likert scale:
0 1 2 3 4 5 6 7 8 9 10
At the conclusion of the vignettes, the participants will also be asked to provide an
overall estimate of guilt in the form of yes or no selection.
Vignette Experiment 2
In this experiment the crime scenario remained mostly the same with the
exception that the victim was not a juvenile and the suspect was not a registered sex
offender. One group received all of the evidence simultaneously. A second group was
provided the evidence sequentially and asked to render an opinion after each item. A third
group received a vignette with a reverse-sequential order of evidence. All of the groups
then rendered a guilt estimate in the form of a yes or no selection. The vignette read:
John Doe has been down on his luck recently and he has been staying with his
brother and sister-in-law, Ron and Jo Doe. John has been living with the family for the
past several months as has had trouble finding a job, due in part to the fact that he is a
convicted felon. Several neighbors come to the Doe home regularly to play cards and
drink alcohol, including Scott and Joanne Smith. Joanne Smith has reported that John
Doe sexually assaulted her, and you have been assigned to investigate the case. Choose
the number that corresponds to your level of suspicion about the report from 0 (not guilty
at all) to 10 (definitely guilty).
Description of Evidence – Sequential and Simultaneous
1. Initial report – You are dispatched to the hospital where you meet Scott
and Joanne Smith. Joanne Smith reports that John Doe raped her the night before at the
home of Ron and Jo Doe, where John resides. Both Scott and Joanne Smith report that
they were intoxicated and do not recall much of the evening. Joanne Smith claims that the
incident occurred in John Doe’s bed and that it was not consensual.
2. SANE nurse report of examination - The SANE nurse’s report indicates
that Joanne Smith stated she had become voluntarily intoxicated the night before and
recalls talking to John Doe in his room, and possibly kissing him. She further stated that
she would not consent to having sex, and therefore must have been raped. Joanne Smith
was given the option of an abortifacient (Day after) pill, which she took. There were no
signs of force or injury noted, however, signs of recent sexual activity were present. A
sexual assault evidence collection kit was performed by the SANE nurse on the victim
and provided to you.
3. Crime scene. Utilizing an alternate light source, body fluid was identified
on the sheets in John Doe’s bedroom. Preliminary lab results indicate the substance may
be semen. The sample was submitted to the lab for results.
4. Victim interview – Joanne Smith states that she and her husband Scott
regularly go to the Doe’s home to play cards and socialize, usually drinking alcohol. On
this occasion, Joanne Smith became intoxicated and does not recall much. Joanne Smith
recalls being in John Doe’s room for some reason and has a vague recollection of kissing
him, but nothing after that. Joanne stated that in the early morning hours she awoke naked
with John Doe, in his bed. She stated that she woke her husband and they went home and
went back to bed. Several hours she awoke and came to realize that she had engaged in
sexual intercourse and told Scott Smith, who then took her to the hospital.
Joanne Smith stated she would not have consented to having sex with John Doe under
any circumstance.
5. Suspect interview – John Doe denies the allegation. John Doe states that
on the evening in question Joanne Smith had been flirting with him, although at the time
he thought nothing of it. John Doe stated that everyone at the home was drinking heavily
and getting intoxicated. John Doe stated that at one point during the evening he went to
use the restroom and when he exited, he was confronted by Joanne Smith, who kissed
him. John Doe states that most of the people had fallen asleep or passed out and that
Joanne Smith came into his room. Before long they were engaged in sexual intercourse.
John Doe stated that although Joanne Smith was intoxicated, she clearly gave her consent,
as she was the instigator of the sexual activity. After awhile they both fell asleep in his
bed. When he awoke, Joanne was not there. John Doe stated that Joanne Smith told him
more than once that Scott Smith must not find out that they sex.
6. Interview of Scott Smith – Scott Smith states that he and Joanne frequent
the Doe’s home to play cards and socialize, and usually drink alcohol. On the night in
question everyone drank more than usual and at some point Scott Smith passed out on the
couch. Scott Smith stated that Joanne woke him up at about 5:00 AM and they went
home and went to bed. At about 10:00 AM they woke up and Joanne became very upset.
She stated that during the previous night John Doe raped her. They got dressed and went
to the hospital where they met a SANE nurse and they contacted the police. Scott Smith
stated that there could be no way Joanne was pregnant because he had a vasectomy.
7. Interview of Jo Doe – Jo Doe states that she and Joanne Smith are friends
and have gone to each other’s home the past to socialize and play cards. Jo stated that
recently Joanne had come to her and asked for information about an abortion. Jane was
concerned she might be pregnant and asked how much an abortion would cost? Jo stated
that she told Joanne that in cases of rape, an abortifacient is free.
8. Follow-up investigation – In a follow-up interview with friends of Joanne
Smith you learn that she had recently become concerned about being pregnant and did
not what to have another child.
A witness from the night of the party tells you learn that Joanne was being
flirtatious with John Doe and was in and out of his room several times during the evening.
This witness states most everyone at the party was aware of Joanne’s behavior except for
Scott, who seemed too drunk to recognize it.
9. Follow-up investigation - In a follow-up interview with Joanne, she
stated she just wanted the whole thing to “go away” and that she no longer wanted to
pursue charges John Doe.
10. Lab report – The DNA from the sheet was matched to Joanne and John
Doe.
Recording Responses – Sequential and Simultaneous
Participants in the sequential group recorded their confidence assessment on a
Likert scale after each item of evidence was presented. Participants in the simultaneous
group recorded a single confidence assessment on a Likert scale after all 10 items of
evidence were presented. The confidence assessment was measured with the following
11-point Likert scale:
0 1 2 3 4 5 6 7 8 9 10
At the conclusion of the vignette, the participants were also asked to provide an
overall estimate of guilt in the form of yes or no selection.
Description of Evidence - Reverse Sequential and Simultaneous
The reverse sequential and simultaneous groups received the same items of
evidence in a different sequence. Some of the verbiage was re-worded for logic and
coherence.
1. Initial report – You are dispatched to the hospital where you meet Scott
and Joanne Smith. Joanne Smith reports that John Doe raped her the night before at the
home of Ron and Jo Doe, where John resides. Both Scott and Joanne Smith report that
they were intoxicated and do not recall much of the evening. Joanne Smith claims that the
incident occurred in John Doe’s bed and that it was not consensual.
2. SANE nurse report of examination - The SANE nurse’s report indicates
that Joanne Smith stated she had become voluntarily intoxicated the night before and
recalls talking to John Doe in his room, and possibly kissing him. She further stated that
she would not consent to having sex, and therefore must have been raped. Joanne Smith
was given the option of an abortifacient (Day after) pill, which she took. There were no
signs of force or injury noted, however, signs of recent sexual activity were present. A
sexual assault evidence collection kit was performed by the SANE nurse on the victim
and provided to you.
3. Crime scene. Utilizing an alternate light source, body fluid was identified
on the sheets in John Doe’s bedroom. Preliminary lab results indicate the substance may
be semen. The sample was submitted to the lab for results.
4. Lab report – The DNA from the sheet was matched to both Joanne Smith
and
John Doe.
5. Witness interviews – In a follow-up interview with friends of Joanne
Smith you learn that she had recently become concerned about being pregnant and did
not what to have another child.
In an interview with a witness from the night of the party, you learn that Joanne
was being flirtatious with John Doe and was in and out of his room several times during
the evening. This witness states most everyone at the party was aware of Joanne’s
behavior except for Scott, who seemed too drunk to recognize it.
6. Interview of Jo Doe – Jo Doe states that she and Joanne Smith are friends
and have gone to each other’s home the past to socialize and play cards. Jo stated that
recently Joanne had come to her and asked for information about an abortion. Jane was
concerned she might be pregnant and asked how much an abortion would cost? Jo stated
that at one point during the course of their conversation, she made the remark that in
cases of rape, an abortifacient is free.
7. Interview of Scott Smith – Scott Smith states that he and Joanne come
regularly to the Doe’s home to play cards and socialize, and usually drink alcohol. On the
night in question everyone drank more than usual and at some point Scott Smith passed
out on the couch. Scott Smith stated that Joanne woke him up at about 5:00 AM and they
went home and went to bed. At about 10:00 AM they woke up and Joanne became very
upset. She stated that during the previous night John Doe raped her. They got dressed and
went to the hospital where they met a SANE nurse and they contacted the police. Scott
Smith stated that there could be no way Joanne was pregnant because he had a vasectomy
8. Suspect interview – John Doe denies the allegation of rape, stating that he
and Joanne had consensual sexual intercourse. John Doe stated that although Joanne
Smith was intoxicated she clearly gave her consent and she was the instigator of the
sexual activity. John Doe states that on the evening in question Joanne Smith had been
flirting with him, although at the time he thought nothing of it. John Doe stated that
everyone at the home was drinking heavily and getting intoxicated. John Doe stated that
at one point during the evening he went to use the restroom in his room and when he
exited, he was confronted by Joanne Smith, who kissed him. John Doe states that most of
the people had fallen asleep or passed out and that Joanne Smith came into his room.
Before long they were engaged in sexual intercourse. After awhile they both fell asleep
in his bed. When he awoke, Joanne was not there. John Doe stated that Joanne Smith told
him more than once that Scott Smith must not find out that they sex.
9. Victim interview – Joanne Smith states that she and her husband Scott
regularly go to the Doe’s home to play cards and socialize, usually drinking alcohol. On
this occasion, Joanne Smith became intoxicated and does not recall much. Joanne Smith
recalls being in John Doe’s room for some reason and has a vague recollection of kissing
him, but nothing after that. Joanne stated that in the early morning hours she awoke
naked, with John Doe in his bed. She stated that she woke her husband and they went
home and went back to bed. Several hours she awoke and came to realize that she had
engaged in sexual intercourse and told Scott Smith, who then took her to the hospital.
Joanne Smith stated she would not have consented to having sex with John Doe under
any circumstance.
10. Follow-up investigation - In a follow-up interview with Joanne, she
stated she just wanted the whole thing to “go away” and that she no longer wanted to
pursue charges John Doe.
Recording Responses – Reverse Sequential and Simultaneous
Participants in the reverse-sequential group also recorded their confidence
assessment on a Likert scale after each item of evidence was presented, while participants
in the simultaneous group recorded a single confidence assessment on a Likert scale after
all 10 items of evidence are presented. The confidence assessments were measured with
the following 11-point Likert scale:
0 1 2 3 4 5 6 7 8 9 10
At the conclusion of the vignettes, the participants were also asked to provide an
overall estimate of guilt in the form of yes or no selection.
Data Collection
Data was collected through SurveyGizmo.com, to administer the questionnaire
and eight surveys. Participants completed an initial demographic questionnaire of design,
collecting general information pertaining to what could reasonably be predicted to be
meaningful and pertinent to be used later to compare to results of the experiments for
analysis. The materials for this study consisted of electronic questions and statements,
and consent form. Participants were provided a unique code that was used in place of
identities, to facilitate anonymity.
Data Analysis
After each item of evidence, participants recorded a confidence assessment of
their belief of the suspect’s guilt/innocence on an 11-point Likert scale from 0 to10. The
results were measured as the difference between the numbers of items (mean evaluation)
and tested against zero for significance. This study used a quasi-experimental, 2 (type of
crime: sex assault, child sex assault) x 3 (assignment: recruit, patrol, investigator) x 3
(evidence presentation: simultaneous, sequential, reverse sequential), between-subjects,
factorial design. The IVs will include duty assignment and evidence presentation order,
while the DV will include a confidence assessment in the level of guilt belief.
Another aspect of this study examined whether emotionally charged evidence
could influence belief in a suspect’s guilt. Participants were advised they were
investigating circumstances surrounding a crime and were receiving several items of
evidence. Both groups received the same vignette of a crime involving a child sexual
assault. One group received evidence sequentially (Appendix A), recording a confidence
assessment of guilt after each item. The other group received the evidence simultaneously
(Appendix B) recording a single confidence assessment of guilt. The predictor variables
were the items of evidence and the criterion variables were the assessments of belief in
guilt.
To evaluate the practical nature of confirmation bias in law enforcement in
criminal investigations it is necessary to consider the tacit knowledge of criminal
investigators and law enforcement officers. Pozzali (2006) reported that the application of
tacit knowledge such as the subtleties and nuances of a criminal investigation are often
biases themselves. Consequently, the order of evidence included a reverse sequence
presentation order, to explore the nature of confirmation bias when exculpatory evidence
precedes the inculpatory evidence. No reverse simultaneous order was utilized because
there is effectively no order of presentation when all of the evidence is visible at one time.
The effect was essentially the same as the existing simultaneous presentation order.
Demographic information collected at the beginning of the study provided basic
information related to individual, organizational, and professional characteristics.
Participants in 3 groups (recruit, patrol, investigations) were assigned to one of the
two conditions (child sexual assault, sexual assault) and one of three orders of
presentation (simultaneous, sequential, reverse sequential). Participants’ global
impression of the case was assessed via one of two (yes/no) guilt measures. Once the data
was collected it was analyzed using SPSS Graduate Pack software.
Ethical Considerations
All of the research participants read and signed an informed consent agreement
stating that their participation in the study was voluntary, that they could withdraw at any
time without any adverse action, and that the researcher was not in a position to influence
their careers. There were no material rewards or incentives provided for participating in
this study. Anonymity was assured by having the participants use a code that was
computer-generated, rather than with their name on the materials. All of the data will be
kept in my possession in a locked file cabinet for 5 years after the study is completed,
after which, the data will be destroyed. The electronic data will be stored on a
freestanding back-up hard drive, and will only be on the researcher’s password-protected
computer. All of the electronic data will be maintained for 5 years, at which time it will
be destroyed.
Summary
This chapter described methods for a quantitative study of the effect of
confirmation bias on criminal investigative decision making. It has been explained that
the use of sworn police officers as participants increased the validity of the research and
that they completed several research instruments. Data analysis included computing
descriptive statistics and analysis of variance to evaluate the relationship between the
predictor variables (bias and emotion) and the criterion variable (level of guilt judgment).
In Chapter 4, the results of the study are presented.
Chapter 4: Results
Introduction
The purpose of this study was to examine the phenomenon of confirmation bias in
criminal investigative decision making among law enforcement officers by utilizing
sworn police officers attending compulsory, annual criminal justice training. This chapter
presents the results of the statistical analysis of the research hypotheses of this
quantitative study. The objective of this study was to examine criminal investigative
decision making by police officers among duty assignments and with extreme emotional
evidence, and to evaluate the responses for evidence of confirmation bias. The
independent variables in this study included the duty assignments (recruit, patrol,
investigations), scenario condition (child & adult), and order of evidence presentation
(sequential, simultaneous, reverse sequential). The dependent variable was a measure of
confirmation bias, in the form guilt belief (0 to 10 scale) among 10 items of evidence.
The research questions were as follows: Does confirmation bias differ among various
assignments (recruit, patrol, investigators) within law enforcement? Can emotionally
charged crimes increase belief in a suspect’s guilt? Does the order of evidence
presentation influence belief in a suspect’s guilt?
In this chapter, I provide a review of the data collected, statistical analysis,
demographic characteristics of the participants, and summarized findings related to the
research questions. Descriptive statistics for participants are followed by descriptive
statistics for the variables used in the study: measures of central tendency for
continuous/interval variables (i.e., means and standard deviations) and frequency
distributions for nominal/categorical variables. ANOVA was used to examine
Hypotheses 1 and 2. A summary of the results is presented at the end of this chapter.
Data Collection
The research was conducted at local police and sheriff’s offices that also serve as
regional training facilities operated by a state department of criminal justice training.
Participants included officers attending mandatory annual training as well as the recruits
at basic training. Additionally, I traveled to the state criminal justice training center to
recruit police trainees who were in basic training. Participants were recruited from six
classes: two for criminal investigators, two for patrol officers, and two for basic trainees.
The basic training recruits were all sworn officers from their respective law enforcement
agencies.
Quantitative data were collected for this study through computer-administrated
surveys using SurveyGizmo.com. Participants were provided with a unique ID number
and password where they could choose to respond in private and complete the survey at
their convenience, in order to assure their anonymity. Officers were told their
participation was voluntary and if they did choose to participate, the access code and
random password granted anonymity. Despite successful pilot testing that validated the
electronic survey process, early in the data collection process there was a problem with
accessibility, and several participants advised they could not access the survey. Access
codes and passwords were reissued and there were no further problems.
Sample Description
The desired sample size was substantially surpassed and the number of
participants exceeded what was proposed in the GPower3 power analysis, therefore the
sample size was large enough to identify statistically significant relationships in ANOVA
and regression analysis. Approximately 200 officers were solicited and 166 officers
responded and completed the survey. Six surveys had been started with no responses
registered. These are most likely attributed to the flawed process that was reported by
participants at the beginning. Of all of the surveys that were completed there were no
discrepancies noted in the data collection process and no volunteers were excluded from
the study.
In the sample, presented in Table 1, 87% of participants were male, 84% had rank
of patrolman (all recruits hold the rank of patrolman), and 78% had a college degree.
Duty assignments were distributed relatively evenly (Recruit n = 55, Patrol n = 54,
Investigations n = 57). A cumulative list of demographic characteristics of the sample is
presented in Appendix L. From an experiential component, the amount of trial and
suppression hearing experience as well as length of service, duty assignment, training
hours and other specialized training hours are presented in Appendix M.
Table 1
Sample Demographic Characteristics
Variable Frequency Percent Valid Cumulative
Percent Percent
Duty assignment
Recruit 55 33.1 33.1 33.1
Patrol 54 32.5 32.5 65.7
Investigations
57 34.3 34.3 100
Gender
Male 145 87.3 87.3 87.3
Female
21 12.7 12.7 100
Education
High school 23 13.9 13.9 13.9
Some college 11 6.6 6.6 20.5
Associated degree 31 18.7 18.7 39.2
Bachelor’s degree 100 60.2 60.2 99.4
Master’s degree 1 .6 .6 100
Model Assumptions and Outliers
ANOVA was chosen in order to study the effect of two or more independent
variables at a time and to assess how the variables interact (Frankfort-Nachmias &
Nachmias, 2008). Several assumptions underlie the univariate ANOVA testing. Foremost
among the assumptions is that the DV in each scenario be normally distributed in the
population for each combination of levels of the between-subjects factors (Field, 2009).
In the analysis of variance models, the samples (duty assignment, scenario condition,
order of evidence presentation) did not vary and no volatility was experienced, thus the
samples were homogeneous. Further, the observations in the study were independent
from one another and the participants did not interact with one another during the survey.
Measures of central tendency demonstrated equal variances among the samples and the
data set did not violate the required assumptions of the statistical tests.
Outliers are scores that are so different from other scores that they can skew the
distribution and alter the statistics (Field, 2009). Outliers were determined by
standardizing the study variables and searching for scores that were in excess of three
standard deviations from the mean. In the present case scores were evenly distributed and
no violating scores were identified.
Relevant Variable Descriptions
Guilt belief. It is important to understand the levels of proof that law enforcement
officers work with because they must be able to articulate their perspective in such a way
that others can comprehend and be persuaded in the same fashion, by the same evidence
(Lyman, 2011). Since no clearly defined standard exists in a specific situation, law
enforcement officers proceed from the lowest measure of guilt judgment known as
reasonable suspicion. Other guilt belief thresholds relevant to this study include probable
cause, preponderance of evidence, and beyond a reasonable doubt.
Reasonable suspicion. This may be understood as more than a hunch or gut
feeling, but less than probable cause. Reasonable suspicion requires articulable facts or
circumstances that give rise to more than a bare, imaginary, or purely conjectural
suspicion (Reynolds, 2013). Reasonable suspicion was described by the U.S. Supreme
Court in Terry v. Ohio (1968) as “the sort of common-sense conclusion about human
behavior upon which practical people...are entitled to rely”. Reasonable suspicion is
depicted as letter B in Figure 1.
Probable cause: Probable cause is the standard required for an officer to justify an
arrest or a search (Black & Nolan, 1993). It consists of more than mere reasonable
suspicion, but less than beyond a reasonable doubt, which is the amount of evidence
required for conviction. Probable cause (C on Figure 1) is the most common level of
proof that law enforcement officers work with because it represents the threshold
necessary for an arrest or search. Probable cause also requires articulable facts in every
arrest or search
Although no constitutional definition of probable cause exists, a series of Supreme
Court rulings ultimately created the language under which probable cause is understood
today (Oliver, 2010). Probable cause may be understood as reasonable grounds for
believing that a specific crime was committed, by a specific person, supported by more
than mere suspicion, but less than prima facie proof. Also known as a slight, well-
grounded suspicion, probable cause must be objectively reasonable and based upon
evidence that reasonable and prudent people would rely (Black & Nolan, 1993). As a
practical matter, an officer who arrives at probable cause may be considered sufficiently
convinced of a suspect’s guilt and is thus convinced for the purpose of this study.
Preponderance of evidence: This is often associated with civil proof where a
conclusion is that a fact at issue is more probably true than not. Another definition is 51%
of the proof (Federal Rules of Evidence). While mostly civil, a preponderance of
evidence provides a solid frame of reference, as it is one of the few guilt thresholds
delineated by law. A preponderance of evidence is depicted as letter D in Figure 1.
Beyond a reasonable doubt: This burden of proof (E on Figure 1) is the threshold
required for a criminal conviction (Lyman, 2011). While it is not beyond all doubt, it
represents the highest threshold on the guilt belief scale. Since no legal definition is
available, the percentage associated with a particular threshold is approximated.
The relevant legal thresholds of guilt belief (and legal proof) on a scale from 0 –
100% are depicted in Figure 1.
A B C D E F
0% 50% 100%
A- Absolute Innocence
B- Reasonable Suspicion
C- Probable Cause
D- Preponderance of Evidence
E- Beyond a Reasonable Doubt
F- Beyond All Doubt
Figure 1. Legal thresholds of guilt
Evidence of Confirmation Bias
This study utilizes responses to survey questions ranging from 0 to 10 to establish
an assessment of guilt belief and a measure of confirmation bias. Figure 2 provides a
frame of reference, demonstrating that 0 corresponds to a belief of absolute innocence
and 10 corresponds to belief of absolute guilt. While this is an approximation, Figure 2
combines the 0 to 10 guilt belief scale with the A-F guilt thresholds. Fifty-one percent
correspond to a preponderance of evidence (C on Figure 2), so scores greater than five
exceed that threshold. Probable cause corresponds to scores between one to five, thus
scores at the higher end of the guilt belief scale (between six to ten) are considered to be
representative of confirmation bias for the purposes of this study. Civil verdicts are made
at the 51% threshold while criminal verdicts are made at beyond a reasonable doubt.
A B C D E F
0 1 2 3 4 5 6 7 8 9 10
A- Absolute Innocence
B - Reasonable Suspicion
C- Probable Cause
D- Preponderance of Evidence
E - Beyond a Reasonable Doubt
F - Beyond All Doubt
Figure 2. Guilt belief scale
A foundational principle of confirmation bias is belief that is unwarranted, or
unsubstantiated by evidence (Nickerson, 2004). In this study, what should be recognized
is that while belief in the suspect’s guilt may be increasing during the reading of the
initial report, there are no corresponding facts or evidence related to the allegation under
consideration, other than the allegation. Thus, if the participant registers an unwarranted
level of guilt belief, confirmation bias is established. Other studies regarding confirmation
bias have similarly established confirmation bias (Ask & Granhag, 2005;
Ask, et al., 2007; O’Brien, 2005; Rassin, 2010; Rassin, Eerland, & Kuijpers, 2010;
Spano, 2005).
Maintaining belief in guilt before considering available evidence or investigating a
case violates basic best practices standards of criminal investigation (Erzinclioglu, 2000;
Lyman, 2011). Further, increasing one’s belief on circumstantial grounds, rather than
following the case facts and evidence to their logical conclusions, is clear evidence of
confirmation bias as the participant is allowing preconceived beliefs and other forms of
bias influence their opinion about a suspect’s guilt. Lastly, participants are aware that
they have ten items of evidence to review, thus, they are well aware further evidence
exists
Emotional Factors
The emotionally charged aspect of this study includes the allegation of the sexual
abuse of a child and is disclosed in the initial report, which is the first question presented
to participants in the child, sequential condition. The emotional characteristic of this
crime is hypothesized to result in a higher initial guilt judgment. Absent corroborating
evidence regarding the actual allegation, this study asserts that such a judgment would
constitute significant confirmation bias as it deviates from investigative protocol and
common sense, with ten items of evidence available for review. The victim statement
discloses that the suspect is a registered sex offender. At this point the participant
recognizes the suspect was once capable of such an act and it is theorized that this will
lead to a greater level of guilt belief. Once established, it is hypothesized that this
confirmation bias exists with sufficient strength to increase suspicion of the suspect and
will manifest through higher scores on the guilt belief scale.
Emotion may work in different directions in law enforcement decision making
(Tiedens & Linton, 2001). In the suspect interview, participants learn that the offense for
which the suspect is registered was consensual sex with his 17-year old girlfriend, when
the suspect was 18-years old. There is significant jury sympathy and reluctance to convict
in this circumstance. The Adam Walsh Act is Federal legislation that creates a
misdemeanor of consensual sexual conduct between juveniles less than 4-years apart, to
avoid mandatory sex offender registration of juveniles. Called the “Romeo & Juliet Act”,
this is a sentencing provision in lieu of offender registry since technically all sexual
conduct under 18-years of age is a criminal offense (Petteruti & Walsh, 2008). As a
result, this is expected to serve as a mitigating factor and may result in a reduction in
suspicion toward the suspect, manifested by lowering of scores on the guilt belief scale in
the suspect interview. Where this is problematic is in the selective search for evidence to
support suspicion (Nickerson, 1998; Frey & Thelen, 2001) to bolster a preferred
hypothesis (Kosnik, 2008) or interpret ambiguous evidence to support a conclusion based
upon preconceived beliefs and expectations and biases rather than case evidence and facts
(Dror, 2005). A key distinction of the confirmation bias phenomenon is that some degree
of guilt belief exists with no clear evidence to support the interpretation (Klayman & Ha,
1987; Kosnik, 2007; Oswald & Grosjean 2004).
As explained in Chapter 3, Pozzali (2006) noted that the application of tacit
knowledge that involves subtleties and nuances related to a criminal investigation, are
often themselves, biases. The order of evidence presentation is therefore important to
consider. Evidence that is reviewed early in a case may be considered with more belief
than is warranted, simply because there is nothing to contradict it. In the sequential
evidence presentation order, inculpatory evidence in the victim interview preceded
exculpatory evidence in the suspect statement. Upon reversing the sequential order of
evidence presentation it was possible to explore the nature of guilt belief when
exculpatory evidence precedes the inculpatory evidence. In the simultaneous condition,
all 10 items of evidence was reviewed before any guilt judgment was made. It was
hypothesized that the simultaneous condition would produce the least amount of
confirmation bias.
Research Questions and Results
Research Question 1
Null Hypothesis 1: Confirmation bias does not differ among assignments.
The first hypothesis was posed to explore the relationship between confirmation
bias and duty assignment, and whether confirmation bias differed among duty
assignments (recruit, patrol, investigations). This was determined by examining the level
of guilt belief in the Initial Report among the duty assignments.
Experimental Design: The design for this experiment included evaluating the
responses to the Initial Report, where the participants learn of the sexual abuse nature of
the complaint. Responses from both the child and adult condition are evaluated and
compared. The initial report variable is first in both the Sequential and Reverse
Sequential evidence presentation orders, and is thus, equally representative of
confirmation bias for the purposes of the study. Recalling that the sample comes from a
law enforcement population, the experimental design is thus, a stratified,
betweensubjects, 3 x 2 x 2 ANOVA. The sample was large enough to permit random
sampling for each condition.
Analysis: A 3 x 2 x 2 ANOVA was conducted to evaluate the effects of three duty
assignments (recruit, patrol, investigations), two scenario conditions (adult, child) and
two evidence presentation orders (sequential, reverse sequential) on guilt belief scores for
the initial report variable. The total frequency, means, and standard deviations of guilt
belief scores from the initial report variable are presented in Table 2.
Table 2 Initial Report Frequency, Means, and Standard
Deviations
N M SD
Child 19 7.68 1.67
Recruit Adult 18 8.17 1.04
Total 37 7.92 1.40
Child 18 6.72 1.90
Patrol Adult 18 5.0 2.43
Total 36 5.86 2.34
Child 18 4.89 2.63
Investigations Adult 20 5.05 2.01
Total 38 4.97 2.30
The analysis revealed a significant main effect for duty assignment (F(2,99) =
22.78, p = .00. There was no significant effect for evidence presentation order (F(1,99) =
1.79, p > .05, or scenario condition (F(1,99) = 0.99, p > .05. Duty assignment did interact
with the evidence presentation order in the expected direction (F(2,99) = 3.35, p. < .05, as
well as scenario condition (F(2,99) = 3.48, p. < .05. Similarly, there was a significant
interaction between evidence presentation order and scenario condition (F(1,99) = 5.07,
p. < .05. The interaction between duty assignment, scenario condition and order of
evidence presentation failed to achieve significance (F(2,99) = 2.04, p. > .05. The results
for the three-way ANOVA are presented in Table 3.
Table 3
Initial Report ANOVA
a R2 = .42 (Adjusted R2 = .36)
Discussion
Recalling that confirmation bias emerges with scores greater than 5, the present
findings revealed that in the initial report variable, confirmation bias is present in each
duty assignment (within .11 and .03 for investigators in Table 2). Belief in the suspect’s
guilt appears to be inversely related to experience, with suspicion greater in recruits (M =
7.92) and least present in investigators (M = 4.97). The interaction between duty
assignment and evidence presentation order, duty assignment and scenario condition, and
evidence presentation order and scenario condition, are relationships that help to provide
some explanation as to what induces change in guilt judgment, although the interaction
may be also be related extraneous nuisance variables. While significant, the effects of
these interactions were small, with the exception of duty assignment (η2 = .37).
Research Question 2
Null Hypothesis 2: Emotionally charged crimes do not increase belief in a
suspect’s guilt.
Sum of Squares df Mean Square F Sig η2
Corrected Model 264.04 11 24.00 6.52 .00 .42
Duty Assignment 167.69 2 83.84 22.78 .00 .32
Scenario Condition 3.65 1 3.65 0.99 .32 .01
Evidence Order 6.59 1 6.59 1.79 .18 .02
Duty Assignment * 25.60 2 12.80 3.48 .04 .07
Scenario Condition
Duty Assignment * Sequence 24.69 2 12.35 3.35 .04 .06
Scenario Condition * Sequence 18.66 1 18.66 5.07 .03 .05
Duty Assignment * 15.02 2 7.51 2.04 .14 .04
Scenario Condition *Sequence
Residual Error 364.40 99 3.68
Total 4955.00 111
Research question 2 evaluated the relationship between emotion and guilt belief.
The initial report included information that the crime involves sexual abuse, in the
respective child and adult scenario categories. In the victim interview participants learn
the suspect is a registered sex offender and in the child condition, that the abuse had been
occurring over a period of time. It was hypothesized that upon learning the suspect is a
registered sex offender there would be an increased belief in guilt, even though no
corresponding factual evidence or case information existed to warrant it, and participants
had been cautioned that ten items of evidence were available to be reviewed.
Experimental Design: This experiment involved a three-way ANOVA that
evaluated the effects of three duty assignments (recruit, patrol, investigations), two
scenario conditions (child, adult), and two evidence presentation orders (sequential,
reverse sequential) on participant guilt belief scores on both the initial report and victim
interview variables.
A baseline of guilt belief was obtained from the initial report (the first item of
evidence) and crime scene (immediately preceding the victim interview). Guilt belief
assessments were then obtained following exposure to emotional stimuli in the victim
interview, and the moderating circumstances in the suspect interview. The child and adult
scenario conditions were compared to examine the effect of emotion related to child
versus adult sexual abuse. Only the sequential order of evidence presentation was
considered because in the reverse sequential condition the victim interview came at the
end, after the suspect interview had the potential to confound the evidence.
Initial Report: Information that the complaint involves a sexual abuse allegation
is contained in the initial report. While the difference among duty assignments was
considered in the first hypothesis, here the focus is on the difference between the adult
and child scenario conditions and the extreme emotional evidence of a child sexual
assault.
Crime Scene: Guilt belief scores from the crime scene variable immediately
precede the Victim Interview and served a baseline for subsequent analyses.
Victim Interview: In the victim interview participants are told specifically of rape
allegations occurring over a long period of time as well as new information that the
suspect is a registered sex offender, a moniker that incites fear in communities
(Levenson, Brannon, Fortney, & Baker, 2007), is referred to as a “disgusting” or
“gruesome” criminal behavior (Capestany & Harris, 2014), and represents an extreme
emotional incident for the purposes of this study.
Suspect Interview: In the suspect interview participants learned of the consensual
nature of the sexual conduct and the mandatory “administrative” nature of the sex
offender registration.
Analysis: An analysis of means and standard deviations revealed the total mean
guilt belief score before the victim interview was M = 6.30 and rose to M = 8.04 after
exposure to the emotional evidence, an increase of M = 1.74. The frequency, means,
standard deviations, and changes in the scenario conditions before and after the victim
interview are presented in Table 4. Of particular interest, the mean guilt belief score
before the victim interview was M = 7.44 for recruits, and M = 9.33 afterwards; an
increase of M = 1.89, registering nearly absolutely guilty on the guilt belief scale.
Table 4 Frequency, Means and Standard Deviations of Crime Scene & Victim
Interview1
Crime Scene Victim Interview Change
N M SD M SD M SD
Recruit 9 7.44 1.74 9.33 .500 +1.89 - 1.24
Patrol 9 6.78 1.48 8.22 1.20 +1.44 - 0.28
Investigations 9 4.67 2.18 6.56 1.51 +1.89 - 0.67
Total 27 6.30 2.13 8.04 1.60 +1.74 - 0.53
1 Child scenario
Initial Report ANOVA: A 3 x 2 x 2 ANOVA was conducted to evaluate the
effects of three conditions on participant’s guilt belief scores in the initial report variable
in one of three duty assignments (recruit, patrol, investigations), two scenario conditions
(adult, child), and two orders of evidence presentation (sequential, reverse sequential).
The results for the ANOVA, as detailed in Table 3, indicated a significant main effect for
duty assignment (F(2,99) = 22.78, p = .00. Main effects were not significant for evidence
presentation order (F(1,99) = 1.79, p > .05, and scenario condition (F(1,99) = 0.99, p
> .05. The analysis revealed significant interactions between duty assignment and
evidence presentation order (F(2,99) = 3.35, p. < .05, duty assignment and scenario
condition
(F(2,99) = 3.48, p. < .05, and evidence presentation order and scenario condition (F(1,99)
= 5.07, p. < .05. There was no significant interaction between duty assignment, scenario
condition and order of evidence presentation (F(2,99) = 2.04, p. > .05. The results of the
analysis are presented in Table 3.
Because of the significant interaction between duty assignment and scenario
condition, I chose to focus on the simple main effects in the duty assignment (Field,
2009). That is, the difference among duty assignments in the child and adult conditions
separately. To control for Type I error across the two simple main effects I set an Alpha
for each at .025. Although the effect sizes were small, there were significant differences
for the child condition (F(2,105) = 9.08, p. = .000, η2 = .15 and the adult condition (F(2,
105) = 14.75, p. = .000, η2 = .22. Univariate Tests data is presented in Table 5.
Table 5
Univariate Tests for Initial Report
Category Sum of Squares df Mean Square F Sig η2
Contrast 74.14 2 37.07 9.08 .000 .147
Child
Error
428.94 105 4.09
Contrast 120.48 2 60.24 14.75 .000 .219
Adult
Error 428.94 105 4.09
Figure 3 presents the simple effects of guilt belief scores in the initial report
variable by scenario condition for duty assignments.
Figure 3. Boxplots of initial report guilt belief mean scores among duty assignment, by scenario
condition
Victim Interview ANOVA: A 3 x 2 x 2 ANOVA was conducted to evaluate the
effects of three conditions on participant’s guilt belief scores in the victim interview
variable in one of three duty assignments (recruit, patrol, investigations), two scenario
conditions (adult, child), and two orders of evidence presentation (sequential, reverse
sequential). The results of the ANOVA revealed a significant interaction for duty
assignment and scenario condition (F(2,99) = 4.40, p. < .05, partial η2 = .082, duty
assignment and order of evidence presentation (F(2,99) = 4.83, p. < .05, partial η2 = .089,
and duty assignment, scenario condition, and order of evidence presentation (F(2,99) =
4.21, p. < .05, partial η2 = .078. No significant interaction was found for scenario
condition and order of evidence presentation (F(1,99) = .84, p. > .05. The findings are
presented in Table 6.
Table 6
Victim Interview 3-way ANOVA
a R2 = .793 (Adjusted R2 = .770)
Discussion
In contrast to the increase in guilt belief in the Victim Interview, the Suspect
Interview variable informs participants that the offense for which the suspect was
compelled to registered as a sex offender was consensual sex with his 17-year old
girlfriend, when the suspect was 18-years old. It is clear that upon learning of the nature
of the offense for which the suspect was required to register, there is a substantial
reduction in guilt belief, as much as 30%. Some scores dropped below where they were
before the victim and suspect interviews were conducted. This finding demonstrates that
the context of evidence is often critical to understanding the totality of circumstances in a
specific case. The change in guilt belief scores is presented in Table 7.
Table 7
Mean Guilt Belief Scores, Victim and Suspect Interviews
Victim Suspect Change Percent
Source SS df M F Sig η2
Corrected Model 1134.33a 11 103.12 34.45 .000 .79
Duty Assignment 171.30 2 85.65 28.61 .000 .366
Scenario Condition 7.25 1 7.25 2.42 .123 .024
Evidence Order 901.02 1 901.02 301.01 .000 .753
Duty Assignment * 26.36 2 13.18 4.40 .015 .082
Scenario Condition
Duty Assignment *
28.94 2 14.47 4.83 .010 .089
Evidence Order
Scenario Condition *
2.50 1 2.50 836 .363 .008
Evidence Order
Duty Assignment *
25.17 2 12.59 4.21 .018 .078
Scenario Condition *
Evidence Order
Error
296.33 99 2.99
Child 9.33 6.22 - 3.11 - 33%
Recruit
Adult 9.25 6.25 - 3.00 - 32%
Child 8.22 5.44 - 2.78 - 34%
Patrol
Adult 8.22 4.89 - 3.33 - 40%
Child 6.56 3.78 - 2.78 - 42%
Investigator
Adult 4.20 2.40 - 1.80 - 43%
Research Question 3
Null Hypothesis 3: Evidence presentation order does not influence guilt belief.
Research Question 3 examined the order in which evidence was presented (or
reviewed) and whether it influenced guilt belief, resulting in confirmation bias. This was
determined by examining guilt judgments after each evidence presentation order
(sequential, simultaneous, reverse sequential). It was hypothesized that when examining
evidence simultaneously, investigators would be less inclined to engage in confirmation
bias.
Experimental Design: This experiment was conducted utilizing data from
Research Question 2. A three-way ANOVA was conducted and evaluated the effects of
three duty assignments (recruit, patrol, investigations), two scenario conditions (child,
adult), and two evidence presentation orders (sequential, reverse sequential) on
participant guilt belief scores in both the initial report and victim interview variables.
Effect of Evidence Sequence: Guilt judgments were registered after each
evidence item was evaluated. Of particular relevance for this hypothesis is the victim
interview. In the sequential order the information about the sex offender registration is
learned before any other evidence is known, allowing the emotional impact to influence
guilt belief. When evidence was presented in reverse order, the overall context resulted
in a different meaning. In the reverse sequential evidence order, the victim interview
comes after all of the evidence is presented and the participant already knows that the
suspect had to register as a juvenile because his girlfriend was not yet 18-years old, not
because he was a sexual predator. The sequential and reverse sequential orders of
evidence are presented in Table 8. Follow-up Investigation 2 includes the victim
declaring she no longer wants to pursue the case, which, because of the nature of the
statement, has too much potential to confound subsequent evidence so it was left at the
end of the sequence in both orders.
Table 8
Order of Evidence Presentation
Sequential Reverse Sequential
1. Initial Report 1. Initial Report
2. SANE Report 2. SANE Report
3. Crime Scene 3. Crime Scene
4. Victim Interview 4. Lab Report
5. Suspect Interview 5. Follow-up Investigation 1
6. Boyfriend Interview 6. School Counselor
7. School Counselor 7. Boyfriend Interview
8. Follow-up Investigation 1 8. Suspect Interview
9. Follow-up Investigation 2 9. Victim Interview
10. Lab Report 10. Follow-up Investigation 2
The effect of reversing the sequence order of evidence substantially decreased
overall guilt. Further, the extreme emotional information revealed in the victim interview
and evidence in the sequential order shown in Figure 4 did not have the same impact on
the same victim interview in the reverse sequential order, as depicted in Figure 5.
Frequencies, mean scores and standard deviations of both scenario conditions in both
evidence presentation orders in the victim interview, are presented in Appendix N.
Figure 4. Child sequential mean scores
Figure 5. Child reverse sequential mean scores Guilt belief was markedly reduced when
evaluating exculpating evidence before being presented with inculpating evidence. In the child
scenario, reverse sequential evidence order, participants knew the suspect had registered as a sex
offender as more of an administrative violation, thus, the anxiety and fear that generally
accompanies registered sex offenders was assuaged by the circumstances (Levenson, Brannon,
Fortney, & Baker, 2007), and the subsequent revelation in the victim interview had the reverse
effect. Thus, the information contained in the suspect interview moderated the effect of the
emotional stimuli.
In the adult scenario condition and reverse sequential evidence presentation order,
participants were aware of the victim’s provocative behavior, corroborating witness
statements, and the suspect’s claim of consensual sex before the victim interview, and
guilt belief was reduced to nearly absolute innocence in the Patrol (M = 0.44) and
investigator (M = 0.90) scores. Changes in means scores between the two scenario
conditions and two evidence presentation orders are presented in Table 9.
Table 9
Changes in Mean Scores of Victim Interview Guilt Belief Scores
Change
Sequential
Percent
Child 9.33 2.30 7.03 -75%
Recruit Adult 9.25 4.10 5.15 -56%
Total 9.29 3.20 6.09 -66%
Child 8.22 2.56 5.66 -69%
Patrol Adult 8.22 0.44 7.78 -95%
Total 8.22 1.50 6.72 -82%
Sequential Reverse
Child 6.56 1.22 5.34 -81%
Investigator Adult 4.20 0.90 3.30 -78%
Total 5.32 1.58 3.75 -70%
In the adult sequential evidence condition there is an experience component
evident in the mean scores. Figure 6 depicts the difference in scores from recruits and
investigators in the victim interview. Recruits were inclined to believe the victim (M =
9.25) whereas investigators were less inclined (M = 4.20). Mean scores dropped even
lower once the suspect interview was considered for both recruits (M = 6.25) and
investigators (M = 2.40).
Figure 6. Adult sequential mean scores
Simultaneous Evidence Presentation: When the evidence was presented
simultaneously and participants reserved judgment until all of the evidence was reviewed,
there was very little guilt judgment in any duty assignment or scenario condition. With
the simultaneous evidence presentation order, participants scored their belief in the
suspect’s guilt once after all of the evidence had been reviewed. As a result, no further
analyses were conducted. In the child scenario the mean score was M = 0.78 and in the
adult scenario the mean score was M = 0.54. Clearly, when reviewing all of the evidence
before rendering a decision on the ultimate issue of guilt, participants were disinclined to
believe in guilt.
Guilt or Innocence: An ultimate determination of guilt or innocence was made
after evaluating evidence in each evidence presentation order. In the child sequential
condition, there were n = 11 guilty and n = 15 innocent while in the adult sequential
condition there were n = 8 guilty and n = 22 innocent. The reverse sequential and
simultaneous evidence presentation orders were mostly innocent. The results are
presented in Table 10.
Table 10
Simultaneous Evidence Presentation Order, Guilt or Innocence
Sequential Reverse Simultaneous
Sequential
Guilty Innocent Guilty Innocent Guilty Innocent
Child 7 2 1 7 0 8
Recruit
Adult 2
6 1 8 0 9
Child 4 5 0 8 0 8
Patrol
Adult 2
7 0 8 0 9
Child 0 8 0 8 0 7
Investigator
Adult 0 9 1 8 0 9
Child 11 15 1 27 0 26
Total Adult 8 22 2 26 0 27
Discussion
The primary purpose of the research question was to determine whether emotional
circumstances can influence criminal investigative decision making. In particular, can
circumstances exist such that emotion replaces evidence as a catalyst for decision making.
Follow-up analyses to the main effect for scenario condition and duty assignment
examined this issue (Salkind & Green, 2011). The follow-up tests consisted of all
pairwise comparisons among three duty assignments. The Tukey HSD procedure was
used to control for Type I error across the pairwise comparisons. The results of this
analysis indicate that in the child scenario, the emotional evidence affected guilt belief
scores. Overall, the 3 x 2 x 2 ANOVA indicates increased guilt belief after exposure to
extreme emotional evidence in the victim interview.
These analyses demonstrate that when exculpatory evidence is considered before
inculpatory evidence, there is less guilt belief toward the suspect. In this scenario guilt
belief was reduced as much as 78% and 95%, in the investigator and patrol categories
respectively (see Table 9). Clearly, the order in which an officer learns about evidence
can be highly influential and when evidence is evaluated simultaneously there is less
confirmation bias.
In sum, these analyses suggest that guilt belief scores were affected by emotion in
both inculpating and exculpating form. Information in the child scenario inculpated the
suspect, whereas the adult scenario included exculpating evidence. Guilt belief scores in
the victim interview demonstrated a much stronger suspicion in the sequential evidence
presentation order. This demonstrates that when a confidence assessment of guilt is made
before all of the available evidence is considered, there is considerable confirmation bias
present since there is no evidence on which suspicion can be substantiated, and
participants were aware that more evidence was available for consideration.
Summary of Results
The foundational question of this study was whether confirmation bias was
different among duty assignments, including police recruits, patrol officers and criminal
investigators. Statistical analysis of research data indicated a significant interaction
between duty assignment and confirmation bias. Secondarily, this study sought to
examine whether extreme emotional evidence could elicit confirmation bias.
When learning of the suspect being a convicted sex offender, we are examining
attitudes, expectations, and biases about whether this affects guilt belief in child
condition. Thus, we examined the level of guilt belief in the child sequential group
following the victim interview, where participants learned the suspect was a registered
sex offender. The mean level of guilt judgment for recruits was M = 8.63, patrol M = 7.11
and investigators M = 5.78. Guilt judgments for the child sequential condition were
similarly disparate. Recruit was M = 1.71, patrol M = 1.85 and investigations M = 1.96.
These data indicate recruits were more likely to believe in guilt, whereas investigators
were less likely, comparatively.
Chapter 5: Discussion, Conclusions, and Recommendations
Introduction
Confirming suspicion is a threat to both the security of the innocent and the
institution of law enforcement. The purpose of this quantitative study was to investigate
confirmation bias in criminal investigative decision making and to address the gap in the
research that exists relative to the lack of law enforcement personnel as research
participants. This study posited three research questions, that confirmation bias may differ
among duty assignments in law enforcement, that emotion can influence criminal
investigative decision making, and that the order of evidence presentation can influence
guilt belief.
This study contributed to the literature related to criminal investigative decision
making and confirmation bias by showing that the effects of confirmation bias differed
among duty assignments and that extreme emotional crimes can influence bias in decision
making. The research design for this study included stratified random sampling. The key
population of participants were sworn police officers attending state sponsored criminal
justice training courses. The research questions examined for this study included the
following:
RQ 1: Does confirmation bias differ among various assignments (recruit, patrol,
investigators) within law enforcement?
H0 1: Confirmation bias does not differ among assignments.
H1 1: Confirmation bias does differ among assignments.
RQ 2: Can emotionally charged crimes increase belief in a suspect’s guilt? H0
2: Emotionally charged crimes do not increase belief in a suspect’s
guilt.
H1 2: Emotionally charged crimes can increase belief in a suspect’s guilt.
RQ 3: Does the order of evidence presentation differ influence belief in suspect’s
guilt?
H0 3: Evidence presentation order does not influence guilt belief.
H13: Evidence presentation order does influence guilt belief.
Overview of Current Study
The goal of the study was to examine decision making by sworn police officers so
the relevance of the findings are generally limited to law enforcement. This study utilized
a sample of 166 police officers in duty assignments ranging from basic recruit to criminal
investigator. Findings were compared using analysis of variance and found that (a)
confirmation bias does exist in law enforcement decision making, (b) emotional evidence
can affect judgment, and (c) the order in which evidence is examined can affect
judgment. This information may lead to greater awareness of the susceptibility of law
enforcement officers to react to evidence of an emotional nature, and of the propensity to
form belief in guilt before considering all available evidence.
Interpretation of Findings
Research Question 1
Research question one examined whether confirmation bias differed among duty
assignments. Once the nature of confirmation bias and guilt belief was established, it was
determined that confirmation bias was greater among less-experienced police recruits. A
three-way ANOVA examined this question and resulted in a finding of significance that
provides confidence in the generalizability of the sample to the population. In addition,
the stratified nature of the sample coming from a pool of law enforcement officers lent
further confidence to the significance.
Research Question 2
Research question 2 examined how extreme emotional evidence can increase
belief in a suspect’s guilt by evaluating responses to the victim interview, and comparing
this condition to one in which the participant learns the suspect is a registered sex
offender. The findings substantiated the research hypothesis that such emotion can
increase guilt belief, however, examination of evidence of a contrary nature also elicited
significant change in guilt belief. In the suspect interview participants learned the suspect
was a sex offender because he had a sexual relationship as a juvenile with his juvenile
girlfriend. As an adult of 18 he was charged with a statutory sexual abuse of his 17-year
old girlfriend, in accordance with the Adam Walsh Act, federal legislation outlining sex
offender registration. The nature of this circumstance was enough to mitigate the guilt
belief, reducing it to where it was before the victim interview.
Guilt judgments in the adult scenario were nearly as strong as those in the child
scenario, which was not expected. Recruits even scored stronger guilt in the adult
scenario (M = 8.17) than in the child scenario (M = 7.68). The emotional effect of a
sexual assault is clearly strong, regardless of a victim’s age or status.
Research Question 3
Research Question 3 examined how the order of evidence presentation can
influence guilt belief. The findings substantiated the research hypothesis that
confirmation bias can exist when evaluating evidence in isolation. Alternatively, when
evidence was evaluated simultaneously (without arriving at a conclusion after reviewing
each item) guilt belief did not arise to the same level it did in the sequential presentation
order. In reverse order, the emotionally charged evidence did not produce the same level
of guilt belief. As described in the previous two paragraphs, when participants had
advance knowledge of mitigating information they were less inclined to believe in the
suspect’s guilt. The order of evidence presentation has significant potential for law
enforcement in the area of reducing confirmation bias.
Summary
Research indicates when conclusions are made about guilt there is a tendency to
engage in seeking confirmatory evidence (Johnson & Fujita, 2012; Klayman & Ha, 1987;
Schrackmann & Oswald, 2014), particularly with evidence is of an emotional nature
(Feigenson & Park, 2006; Weiner, et al., 2006). As Festinger (1957) noted, when a
decision maker rejects one attractive thing over another, it is not easy to compensate for
and there is strong post-decisional dissonance, often resulting in a seize and freeze
mentality. When this occurs, emphasis is placed predominantly on preexisting cues,
rather than later information (Kruglanski & Webster, 1996).
Wiener, et al. (2006) explained that emotions legal decision makers anticipate
when contemplating their conclusions could mediate the effect of evidence on their final
judgment. Research also shows that the influence of anticipated emotion can override
rational decision making (Shiv & Fedorikhin, 2002). Lastly, it should be understood that
while a search for evidence may be complex, often the review and analysis of it is not.
This is critically relevant in an investigative context where confirmation bias can result in
decisions being made based upon evidence gathered early on, rather than the totality body
of evidence; a proposition supported by this study.
Summary of Implications
Recalling confirmation bias in this study was measured by the increase in guilt
belief across survey questions. When participants learned the subject in the child scenario
was a registered sex offender, their belief in his guilt rose without any evidence to support
it. This constitutes seeking or considering confirmatory evidence to confirm a hypothesis
of guilt (Rassin, et al., 2010). A more objective means of investigation would be to wait
and see what the totality of the evidence shows.
Biased assimilation of evidence occurs when the information is considered more
favorably if it supports initial attitudes (Lord, et al., 1979; Masnick & Zimmerman,
2009). This was a likely reason why guilt belief increased after learning of the suspect’s
registered sex offender status, and was subsequently reduced when learning of the
consensual circumstances requiring the suspect to register as a sex offender. Initial
attitudes toward both sets of circumstances were strong enough to overwhelm restraint,
and register a strong belief. The compelling nature of these beliefs cannot be overstated,
as they appeal on an emotional level and are highly influential on decision making, even
on the part of criminal investigators.
Attitudes, emotions and prior beliefs are strong factors in confirmation bias as
they generally involve deeply rooted beliefs. For instance, belief about a suspect’s guilt
may be the result of an overzealous desire to protect children or skepticism about the
recidivist nature of registered sex offenders. Emotion can bond an opinion to an attitude
or belief in such a way that it becomes an unyielding position. These findings lead to the
conclusion that confirmation bias is not only present in criminal investigative decision
making but it can be systematically problematic if not recognized and contained. The
dynamic nature of evidence requires context to be properly understood. While on its face,
evidence may appear to lean one way or the other, but context is required to fully
understand the totality of circumstances, and avoid erroneous decision making.
Limitations of the Study
Construct validity of the survey is subject to the honesty and certainty of the
participant’s responses (Creswell, 2003; Frankfort-Nachmias & Nachmias, 2008). If
participants under or over-estimated their beliefs, the reliability of the survey may be
compromised. Initially, demand characteristics are of concern given the familiar nature of
the evidence by the participants. The background information revealed in the implied
consent form together with the incident summary gives an idea of what the case entails
and any experienced officer is familiar with the circumstances presented in this study.
After all, the scenarios in this study come from this researcher’s own experience as a
detective. Thus, participants could very well be guessing at what they anticipate is the
proverbial rest of the story that so often accompanies complaints of this nature. Since any
explanation about bias in the study would most likely have caused more confusion than
clarity, participants were simply told that the study involves law enforcement decision
making, and they should answer as honestly and accurately as possible.
Further, when participants registered their belief in the suspect’s guilt in the
respective vignettes, they very well may have estimated, rather than evaluated, their
responses compared to the specific circumstance. Various biases such as response bias,
social desirability bias may have had a confounding affect on the survey, limiting the
validity of the findings, however, surveys are commonly used and are identified as great
sources of information (Creswell, 2003).
These limitations are an intrinsic part of survey research, not just the present
study. Electronic data collection was designed to maintain anonymity in order to mitigate
this as much as possible. Additionally, during participant recruitment officers were
advised of the research gap created by the use of non-law enforcement personnel in police
related research and were encouraged to answer all questions honestly and accurately.
Stereotyping may have been influential while evaluating the guilt of a registered
sex offender accused of another child sex offense. If the notion, he’s done it before so he
is capable of it again, were to invade a participant’s deliberation, stereotyping could
overwhelm practicable recidivism rates, affecting the reliability of their conclusion. So
authentic is this threat that it is impermissible for jurors to be made aware of a
defendant’s prior criminal history during guilt deliberations. Only if a verdict of guilt is
returned, are jurors permitted to learn about a defendant’s prior bad acts (FRE 404b)
during sentencing deliberations. In spite of these potential limitations, the surveys utilized
in this study are common sources of information and should not exclude the substance of
the research findings.
The availability heuristic explains that people tend to answer questions of
probability by summoning examples that come readily to mind (Kahneman, 2011;
Tversky, & Kahneman, 1973). Clearly, actual examples of criminal conduct are more
frequently in the mind of police officers than college age research participants, thus their
responses are more realistic. Unfortunately, there is no greater influence on what a person
can call to mind than entertainment and television in this day and age of law enforcement
programming. While a law enforcement officer may compare a scenario to a case they
worked last week, a student is more likely to imagine it according to what they saw on a
screen.
As a result, this study was designed with law enforcement population in mind.
Given the trend in similar research utilizing convenience sampling for research
participants, this study sought to fill the existing research gap by utilizing law
enforcement officers so the findings may be more readily, and accurately generalized.
While being more relevant to law enforcement officers, the findings are obviously limited
to that population.
Social Change Implications
The clearest social change implication is the connection between confirmation
bias and wrongful conviction. A consistent presence in wrongful conviction exonerations,
confirmation bias is of crucial importance in both forensic and legal contexts.
Recognizing the construct and its potential for error is a matter of education and
awareness, however, implementation of measures to lessen and prevent it may be more
challenging. The fact that it may be challenging, however, should not be a deterrent, but a
call to action for those who operate at the intersection of law and psychology.
The present study offers clear evidence of how confirmation bias tends to emerge
in criminal investigation and provides several practical suggestions on how to avert these
situations. Any research that can improve on law enforcement decision making has
potential for social change. This study can help by disseminating research findings within
the law enforcement community first to acknowledge the ubiquity of criminal
investigative confirmation bias in wrongful conviction cases from an empirical
perspective, and secondly to understand the practical importance of recognizing and
limiting confirmation bias in the practice of investigative decision making. By including
law enforcement officers as participants and law enforcement constructs in the process,
these research findings have the potential to be received auspiciously by the law
enforcement community and therefore, to more successful in effecting social change.
Recommendations for Action
The first recommendation for action is the continued, if not increased, inclusion of
law enforcement personnel as research participants in order to investigate psychological
constructs in a forensic context. Not only can this bring practicality to the subject matter,
it can advance forensic related research toward mainstream academic discussion. The use
of 18-year old college freshman for experimentation and the subsequent generalization to
professional law enforcement may be convenient from a research perspective, but it is
wholly impractical from a realistic standpoint. Recommendations are not likely to be
considered when presented in academic jargon and directed at law enforcement, rather
then enlisting their input and including them in the process from start to finish.
Secondly, greater cooperation between law enforcement and academic researchers
can facilitate the exchange of practical findings. Eliminating layers of bureaucracy can
help to implement positive suggestions more directly and efficiently, teaching and
training the actual participants who work on the street. Establishing positive peer-to-peer
relationships may even result in feedback from the user perspective. Imagine relaying
findings on interview techniques directly to police interviewers, encouraging them with
information and instruction. The implications of such collaboration is encouraging and
potentially constructive
The most important recommendation this research can make is for investigators to
withhold judgment until collecting as much evidence as possible, and focus more on what
is known than what is suspected. Conscious awareness of the potential for bias can help to
avoid premature judgments. More importantly, from a law enforcement perspective
focusing on what is known instead of suspected can help to enhance the prosecutive merit
of a case by eliminating speculation and potential avenues of defense. Ultimately,
convicting the guilty and avoiding wrongful accusations and convictions has potential to
avoid negative impacts that cannot be underestimated. Deliberating on conclusive
evidence can help to remove doubt and ultimately, enhance community relations by
engendering trust and confidence in law enforcement. Including appropriate training in
the basic training and in-service training curriculums can be an effective means to achieve
this recommendation.
Psychologists and law enforcement need to collaborate to develop better decision
making skills and reduce the presence of confirmation bias in wrongful conviction.
Practical recommendations can be made in layman terms that show the way forward in
recognizing confirmation bias and reserving personal judgment until an entire body of
evidence is considered in a case.
Investigators must implement judgment, not simply common sense or some sort of
undefined rule; judgment is mental activity that draws upon beliefs, emotions, sensations,
and involves integrating new information into existing patterns of thought, readjusting
patterns to accommodate new perspectives where appropriate (Garsten, 2006). Judgment
is affected by life experience, personal beliefs, and implicit professional familiarity, all of
which can influence decision making. Exercising temperance in judgment can increase
the criminal investigator’s effectiveness and awareness is an important first step, however
responsibility for accountability must emanate from the parent agency.
Acknowledging that confirmation bias is real and seeking information that is
contrary to a favored hypothesis is an unnatural process so it takes a conscious effort.
Investigators in particular should limit overconfidence and seek out consultation
regarding their investigative findings, such as an informal verification of investigative
conclusions through the use of a peer or supervisor. It is a professional responsibility to
challenge findings in order to ensure truthfulness and accuracy to avoid convicting the
innocent, and to fortify weak elements in a criminal case in order to convict the guilty.
The phenomenon of confirmation bias in law enforcement is a subject that merits
discussion in public forums as community policing initiatives. In the next few years, the
U.S. Department of Justice’s Community Oriented Policing Services (COPS) will
emphasize reforming policing, aligning organizational management, and pursuing
proactive problem solving, preventing biased-based policing, and bringing greater
resources to openness, transparency, and identification of problem solving strategies.
Right or wrong, contemporary policing is in the public spotlight now more than ever, and
there is no time like the present for a stronger, sustained relationship between law
enforcement and psychology practitioners to partner for the purpose of addressing issues
that affect the public.
Future Research Implications
This study showed that confirmation bias is related to restraint in judgment, that it
differs among duty assignments, and that extreme emotional evidence and evidence
presentation order can influence guilt judgment in law enforcement officers, affirming
Nickerson’s (2004) observation of the ubiquitous nature of the phenomenon. Linking
these variables with confirmation bias attempted to bridge the gap between research in
bias and decision making, while using a law enforcement research population.
A recommendation for future research that corresponds with what Eerland and
Rassin (2012) suggested deals with the effect of crime severity on the evaluation of
evidence. Clearly the emotionally charged crime of child abuse proved sufficient to
heighten guilt judgment, but other crimes may also stimulate strong emotional such as
elder abuse or domestic violence. In addition, there is a tendency in criminal investigation
to take severe crimes more seriously, particularly when there is more active supervision
by administrators or when the media are involved. This goes hand in hand with the study
of extreme emotion. As this research has shown, when investigators are confronted with
extreme emotional crimes, they are susceptible to the same inducement as others to react
or respond, when they should be collecting facts and information and withholding
judgment until the facts are in. The results of this study indicate that more research is
necessary to understand the influence of emotion experienced by investigators at the time
of decision making, also recommended by Wiener, et al. (2006).
Summary
“The ability to objectively evaluate information is a hallmark of rational decision
making” (Rassin & Judge, 2007). It is undisputable that bias is ubiquitous in the human
experience. To deny the presence of bias is a certain plan for failure in one’s professional
endeavors. In my own military training I learned that you either plan to succeed, or plan
to fail; a philosophy that would serve me well in a 20-year law enforcement career.
Accounting for bias, recognizing it’s presence and being honest about it’s effect, can have
positive implications, and prepares the legal practitioner to deal with issues before they
arise. This is not a case-by-case situation or a one-time allowance; it is a form of behavior
and a plan for success. Objectivity, skepticism, and accountability are not only a means
for successful investigative decision making, they are a mechanism to preserve individual
liberty.
It is unlikely that bias will ever be eradicated from criminal investigative decision
making. Despite federal, state and local laws, Rules of Evidence, Rules of Criminal
Procedure, as well as numerous local ordinances, policies and procedures governing
police procedure, confirmation bias continues to exist in decision making. As an inherent
part of the human psyche, bias tends to get lost in the process and often becomes the
source of legal argument. Further, judgment emerges from personal dispositions,
emotions, beliefs, and tacit knowledge, and can be biased by factors that are not
connected to specific facts and circumstances. Any exercise involving judgment has the
potential for human error as a consequence, however, confirmation bias need not be as
compulsory a consequence as it often is. For example, upon interviewing a victim, an
investigator believing or disbelieving their story may proceed to seek evidence to confirm
their suspicion. This is an unnecessary initiation of confirmation bias.
Nothing in this study is intended to suggest that judgment be eliminated or
replaced from decision making by police officers. On the contrary, mixing skilled
investigative acumen with the right dispositions or habits of affective responses should be
cultivated and tempered with prudence. This skill requires the decision maker to be on
guard against allowing personal beliefs from intruding into matters requiring professional
judgment. Beliefs are imperfect, whereas facts more accurately represent of what may or
may not have occurred and are, in theory, impervious to bias. Ultimately a belief is only
as reliable as the facts that support it. Judgment is the very tool with which the decision
maker can identify and limit bias in decision making.
Reaching beyond the research and statistical analysis of the data, the core issue
continues to be fairly simple; criminal investigators must reject the temptation to profile
an offender while in the fact gathering process of an investigation, and endeavor to
reserve judgment until all of the facts are in, considering alternative scenarios as critically
as others. Confirmation bias often leads to case building instead of fact gathering, where
the investigator filters their perspective of the evidence in a case through the lens
provided by a conclusion of guilt. As Findlay and Ross (2006) noted, supporting evidence
is elevated in significance and viewed as consistent with the other evidence, where it is
then deemed relevant and supportive of the main conclusion of guilt. Vigorous skepticism
of a favored theory is a strong guarantee against investigative tunnel vision, premature
theory development, attribution error and ultimately, confirmation bias. Tenets of fairness
and justice demand no less.
