Philosophy

John Doe

Florida International University

This paper is much longer than yours needs to be since I added additional analysis / tables beyond those required. I wanted to make sure you saw multiple statistical tests and their write-ups.

Research shows that blame assessment is influenced by the presence of consensus, and this effect intensifies when the accused is a male. However, limited amounts of studies observe these assessments of blame on social media. We analyzed how cheating on an exam was perceived when a consensus had been established that either supported or opposed the cheating (study one) and when the gender of the cheater was manipulated (study two) on a social media platform. In both studies, participants read the Facebook post of a student who confessed to having cheated on an exam followed by a set of eight comments that varied in their level of support (all supported the cheating, all opposed the cheating, or some supported it while others opposed it). In study one, the gender of the cheater was female. In study two, gender was manipulated with participants randomly assigned to either a male or a female condition. Participants assessed how wrong and immoral the cheating was in study one and how wrong and appropriate the cheating was in study two. In general, participants assigned to supportive comments condition agreed that the behavior being not wrong. Participants assigned a female character also rated her behavior as more appropriate and less wrong than their counterparts assigned to a male character. These results suggest that wrongful behavior is more acceptable in the presence of consensus, and when the defendant is a female.

Keywords: blame assessment, assessment of blame on social media, assessment of blame and gender, effect of gender on perception of blame, assessment of blame and consensus

The Influence of Digital Consensus and Gender: Conformity and Moral Decisions

The rise of technology has altered every aspect of American life, especially in social interactions. Consumers now look to each other online to provide reviews on products and consensus on the validity of marketing claims before they make a purchase. While some head to social media to enjoy comments on the latest hot topics people also use social media sites for more than conversation. Social media sites now compete with traditional avenues, such as television or newspapers, as a primary source of information on important subjects ranging from headlining news to political commentary (Winter et al., 2015). It was once thought that the internet would not possess the persuasive power of real face to face encounters with people (Kelly et al., 2017). However, growing research has shown that computer screens cannot protect the mind, or corresponding behavior, from social influence (Kelly et al., 2017). Researchers are now interested in learning to what extent social media comments affect the thoughts and behaviors of others. This paper first reviews relevant social media research then describes two studies concerning the social influence of consensus and gender on matters of conformity while making moral judgments.

Social influence is a change in thinking or behavior caused by a person’s perception of how they are viewed by other people in society (Cohen & Golden, 1972). This influence can produce a set of social “norms,” which can be explained as a set of thoughts or behaviors which society, at large, deems acceptable. One type of social influence is called conformity. The theory of conformity supposes that people will change the way they think or behave to align with social norms. Researchers believe that the influence behind conforming can be either informational or normative (Cohen & Golden, 1972). Informational influence motivates a person to change their opinion because they received new information. For instance, when a person is faced with uncertainty on who to vote for in an election, their uncertainty will heighten their need to be influenced by any information offered to them by others. In contrast, the normative influence has more to do with fitting in and being accepted, this is exemplified by changing your supportive position on a candidate after hearing that all of your friends support his opponent (Cohen & Golden, 1972). It will be interesting to know if the same effects can occur online.

This research aims to investigate factors for changes in behavior, it will address the concept that social influence can cognitively affect various types of decisions. The concept is plausible on a level of consumerism; thus, companies pay millions of dollars on advertisements and marketing campaigns to sway opinions and sell products. However, will the concept be applicable to moral decisions? In fact, researchers found that participants primed by moral reasoning can quickly alter the parameters of what they consider to be moral, without justification (Kelly et al.,2017). The study first primed a group of participants with a high percentage of consensus before asking them to agree or condemn the actions taken during a moral dilemma. Researchers found that participants rated actions more condemnable when the consensus condemned the actions. The same paper presents a replication with extension study in which researchers primed participants with either logical reasoning or emotionally charged statements about the moral dilemma. They found that emotionally charged priming using words like “barbaric” and “horrible” had a higher consensus effect than the use of statements like “did not have the right” which embodied logical reasoning (Kelly et al.,2017). Accordingly, one can expect this phenomenon to occur on social media.

According to research, social media platforms can alter viewer ideology and behavior by presenting information in a meme like fashion (Winter et al., 2015). The initial headlines, pictures, likes, and comments presented on a subject can impact the readers opinion; especially when the persuasion is negative (Winter et al., 2015). This seems to be in line with the exemplification theory, which is a model of the way humans’ group, store, and recall information. Related information, such as the awareness of sharks and beaches, are stored together so that the complete set can be retrieved when making a decision (Spence et al., 2017). Through cognitive devices, people make judgments that are influenced by the first set of information that comes to mind, regardless of its logical strength (Spence et al., 2017). For example, these devices are responsible for the heightened fear of sharks one may experience at the beach after engaging in a shark week movie marathon even if sharks have never been spotted on that beach. Hence, at any given moment, people may make decisions based on the set of information that reaches their brain first, ignoring better information that may not have made it on time to be processed into the decision (Spence et al., 2017). This explanation suggests that social media comments could have a priming effect on the morality of a situation, and a person may use the comment section to make an instantaneous moral decision in line with the consensus instead of using all of the applicable information stored within their memory. Accordingly, a participant can be propagated to respond in agreement with a moral consensus in condemning an action; even though the person would not usually feel that the specific action taken was wrong. In addition, the desire to conform to others can influence behavior even when a person is aware that a behavior is wrong.

In 1951, a revealing consensus experiment was conducted (Asch, 1951, as cited by Kundu & Cummings, 2013). Real participants were placed in a room with four other acting participants. They were shown a series of lines and asked to match the correct line to a given sample with a verbal response. When the four actors in the room repeatedly chose the incorrect answer, participants agreed with the actors more than 30% of the time. However, in the control group, where participants answered their questions alone, they chose the wrong line less than 1% of the time. The distinction here was clear. The contrast highlights the effects of social influence (Kundu & Cummins, 2013). Participants in the experimental condition knowingly chose wrong answers solely to fit in with the social norm of that room. These studies have been replicated many times to confirm their findings: people tend to conform to social norms (Kundu & Cummins, 2013). In a replication of the aforementioned experiment, researchers used a series of vignettes, instead of lines, to see whether they would get the same results using a moral judgment theme. Results revealed that moral judgments could be substantially swayed with a consensus of peers who find immoral acts permissible (Kundu & Cummins, 2013). Conversely, people could also be swayed to view moral acts as impermissible. Accordingly, both the way information is presented, and the appearance of consensus can play a role in conformity.

This present study sought to determine the extent to which priming influenced participant consensus on the ethical nature of a dilemma. Participants were given a Facebook profile belonging to a student (Abigail) who mistakenly received the answer key to her exam. The test was expected to be curved. However, because she used the answer key for a perfect score, the test was not curved, and all the other students failed the exam. Abigail presented this dilemma on a Facebook post asking for direction on what would be the best moral action to take. In the supportive condition, participants received a copy of the post that included supportive feedback from Abigail’s friends encouraging her to remain silent. In the opposed condition, participants were given a copy of the post that included oppositional comments while encouraging her to turn herself in. The final mixed condition includes feedback that was equally supportive and oppositional.

In general, it was predicted that participants who read unanimously supportive feedback would rate the Facebook user’s conduct as more acceptable than participants who read unanimously oppositional feedback, with those who read mixed feedback falling between these extremes.

More specifically, participants in the unanimously supportive condition would more strongly agree with supportive survey statements (“Abigail’s behavior was understandable, “Abigail’s behavior was reasonable”, “Abigail’s behavior was appropriate”, “I would advise Abigail to keep silent”, and “I would try to comfort Abigail”) and more strongly disagree with oppositional survey statements (“Abigail’s behavior was wrong”, “Abigail’s behavior was unethical”, “Abigail’s behavior was immoral”, and “Abigail’s behavior was unacceptable”) compared to participants in the unanimously oppositional condition. Participants in the mixed condition were expected to fall between these extremes. However, participants in both the unanimously supportive and unanimously oppositional conditions were expected to strongly agree that they would give Abigail the same advice that her friends gave her.

Methods

Participants

In 2020, 140 people at Florida International University were randomly solicited to participate in a study. Of these participants, 39.3% ( n = 55) were male, 57.9% ( n = 81) were female, and 2.9% were unspecified ( n = 4). Participant age ranged from 16 to 59 years with the average participant age being M = 23.86 ( SD = 7.15). Of this sample population, 43.6% consisted of Latino/a ( n = 61), 16.4% were Black ( n = 23), 22.9% White ( n = 32), 6.4% Asian ( n = 9), 2.1% Indigenous ( n = 3), and 8.6% classified themselves as other ( n = 12). See Table 1.

Table 1

Study One Demographics

Materials and Procedure

After approaching persons on the university campus, researchers explained the potential risk and benefits of the survey. If the potential participant gave verbal consent to join the study, then he or she was given instruction along with one of three documents that consisted of a Facebook post and a four- part questionnaire. Participants were introduced to the Facebook page of a woman who wanted feedback on a moral dilemma she was facing. Abigail Foster mistakenly received the answer key to her final exam from her professor as he handed out test forms. Instead of giving the key back to her professor, Abigail used the key on her exam and received a perfect score. Because of Abigail’s decision, the test was not curved, and all the other students failed. The Facebook documents used in each condition were identical in nature, apart from the comments section of the post. The comment section portrayed three different consensus conditions (opposed, supportive, or mixed) exemplified by the type of comments left by Abigail’s Facebook friends.

In the first part of the study, participants were exposed to one of three conditions in the form of a printed document that resembled a Facebook profile page complete with Abigail’s name, cover photos, friends section thumbnails, advertisements, and an about section reveling her corky personality. Her post explains her dilemma and below are 8 comments. The comments were from the same people in all three conditions and only slight alterations were made to the comments between conditions. The first condition (support), revealed a consensus of comments that praised Abigail for taking advantage of this opportunity (don’t feel bad, you got lucky…) The second condition (opposed) was a consensus of comments that showed disapproval for the action she took and encouraged her to remedy the situation (you can’t take the grade, your integrity…). In the third condition (mixed) there were four supportive and four opposed comments with no clear consensus (don’t feel bad, you got lucky, you can’t take the grade, your integrity…).

After reviewing part one (Facebook page and comments), participants continued to the survey portion of the study with instruction not to look back at the previous page (to protect the integrity of the attention check question in part five). Part two consisted of 7 statements to be answered by a six-point Likert scale. Participants were asked to rate their impressions of Abigail’s behavior (1 = strongly disagree, 6 = strongly agree) through statements that began with “Abigail’s behavior was…” and ended with 7 different words which showed support or opposition in a randomized order. The words chosen to show a consensus of support were understandable, reasonable, and appropriate. The words chosen to display opposition were wrong, unethical, immoral, and unacceptable. Part three of the study similarly used a six-point Likert scale (1 = strongly disagree, 6 = strongly agree); this portion presented the participant with the opportunity to rate how they would advise Abigail (I would advise her to keep silent, I would comfort her, I would give her the same advice as her friends), how they would have behaved in the same situation (I would keep silent, I would confess), and their impression of Abigail as a person (warm, good natured, confident, competitive, sincere, moral, competent). In part 4 of the study, participants provided demographic information such as, gender, age, ethnicity, primary language, relationship status and whether the person was a current FIU student. The final part of the survey was an attention check to see if the participant noticed what condition they were exposed to. Participants were asked to select the correct option without looking back to the post (Feedback supported her behavior, Feedback opposed her behavior, Feedback was mixed). Subsequently, researchers concluded the survey with a debriefing reveling the nature of the study and the main hypothesis along with an explanation on how the subject’s participation aided the study.

In this study there were over 20 dependent variables; however, many were ignored for this analysis. The main focus was on whether participants passed the manipulation check, whether they agreed with the perceived consensus in their condition, and how they responded to a particular statement in part two. Researchers hypothesized that the manipulation would be noticed. In addition, they hypothesized that people in the support and opposed conditions would agree to consensus by more strongly agreeing to the statement: I would give Abigail the same advice that her friends gave to her. Finally, researchers predicted that support condition participants would more strongly disagree and opposed participants would more strongly agree with the statement “Abigail’s behavior was wrong” while mixed condition participants would have varied results.

Results Study One

The first analysis done in this study was conducted on the manipulation check. Researchers ran a chi square test of independence using condition as an independent variable (support, oppose, mixed) and feedback selection as the dependent variable (supportive feedback, opposed feedback, mixed feedback) to determine whether participants in each condition were able to correctly identify their condition; the results were significant , X2(4) = 129.03, p < .001. Most participants in the supportive condition were able to correctly identify their conditions feedback (75.6%). Most opposed condition participants were able to correctly identify their conditions feedback (76.7%). Most mixed condition participants were able to correctly identify their conditions feedback (81.3%). Cramer V was large, .689. This indicates that the participants were aware of the presented consensus within their condition. See Table 2.

Table 2

Crosstabs and Chi Square – Study One

The main analysis for this study is related to whether participants in the support and opposed conditions differed in their agreement with the presented consensus. For this analysis, researchers conducted an independent samples t-Test with the support and oppose conditions only for the independent variable and “I would give Abigail the same advice that her friends gave her” as the dependent variable, which was not significant, t(89) = 0.16, p = .164. Participants did not differ in the support condition ( M = 4.24, SD = 0.77) and the oppose condition ( M = 4.27, SD = 0.84). These results indicate that support and oppose groups did not differ in their level of agreement with their respective consensus. We omitted the “Mixed” condition here since the question would be confusing for the participant (Abigail’s’ friends gave her conflicting advice in this condition, so the “same advice” wording in the question is not applicable in that condition). See Table 3.

Table 3

t-Test “Same advice as friends”- Study One

Table 4

ANOVA – “Abigail’s behavior was wrong” – Study One

As a final analysis, researchers ran a One-Way ANOVA with condition as the independent variable (support, oppose, mixed) and agreement with the statement “Abigail’s behavior was immoral” as the dependent variable, which was not significant, F(2, 135) = 2.98, p = .054. Although close to significant, participants had similar levels of agreement with the statement more in the support condition ( M = 3.43, SD = 0.98), the oppose condition ( M = 3.86, SD = 0.91) and the mixed conditions ( M = 3.81, SD = 0.87). Since the test was not significant there was no need for a post hoc test. See Table 5.

Table 5

ANOVA – “Abigail’s behavior was immoral” – (Study One)

Researchers initially predicted that participants would notice the studies manipulation in all three conditions. Results support this hypothesis. It was also predicted that participants in the unanimously supportive condition would more strongly disagree with the statement “Abigail’s behavior was wrong” than participants in the unanimously oppositional condition, with participants in the mixed condition falling between these extremes. This hypothesis was partially supported by analysis as participants in the supportive condition disagreed with the statement more that participants in the opposed condition. However, there was no difference between the opposed and mixed conditions indicating that the IV effect may not have been as strong as predicted for the opposed condition. The final prediction was that participants in the supportive and mixed conditions would both agree to give Abigail the same advice that her friends gave to her. Results support this hypothesis; however, neither condition portrayed a mean close to the strongly agree score on the Likert scale. While most participants on average may agree, they do not very strongly agree with the statement. It is possible that results were slightly hindered by word choice in our survey questions. After all, there was no difference when it came to our “immoral” question. Thus a behavior may be seen as “wrong” but not raise to the level of being “immoral”. It is also conceivable that college student participants were slightly biased in their responses due to the nature of the study (cheating on a college exam) while on school grounds. Additionally, it would be interesting to note if moral consensus would fluctuate in any way as a response to the gender of the student who sought out the consensus; this will be examined in study two.

Study Two

Gender is defined as male or female human, and in this paper it will be used interchangeably with sex (Dictionary.com, 2020). Gender has been a widely researched topic within psychology. Moreover, the ways that gender can diverge, or overlap, has become a hot topic in headlines due to many cultural and legal issues (Hingston, 2013). Unconscious gendered bias is widely blamed for discrimination in employment and pay. Gendered stereotypes are now a topic at the forefront of college campuses across the nation in an attempt to redefine social norms. Accordingly, it will be interesting to investigate whether gender plays a role in moral consensus.

Widely accepted views of gender normality begin formation during childhood (Signorella & Frieze, 2008). Again, in line with exemplification theory, people use gendered schemas to store information for each sex (Spence et al., 2017; Signorella & Frieze, 2008). As society reinforces these gender norms, cognitive devices (called heuristics) are formed to allow the mind to bypass all available information and take a mental shortcut when making quick judgments (Signorella & Frieze, 2008). In order to understand how these heuristics can alter study participant perception, one must understand gendered schemas as applicable to education and criminality.

In a research study conducted at Riverside College, data was generated from sixty-four students enrolled in a variety of academic disciplines (Francis et al., 2014). The study involved interviews, focus groups with teachers, seminar observation, and student discovery groups. Researchers received overwhelming reports that there were no true differences between the sexes in general or at the college. However, as conversations ensued, researchers also discovered that both male and female subjects had similar thoughts as to gendered social norms when speaking of personal life experiences. Males were perceived as highly intelligent yet lazy. Conversely, women were viewed as hard workers who are diligent about the effort placed into their assignments; however, they were not specifically labeled as unintelligent. After reflecting, many students in the study still dismissed the reported differences as a byproduct of individual choice. Nonetheless, their individual preconceived notions based on life experience were congruent with common gender stereotypes. This finding implies that within a scholastic setting, participants of a study may view woman as hard working and men as lazy. This finding would also suggest that a behavior like cheating could be perceived as a deviation from a woman’s gendered norm. To explore this concept it seems necessary to examine how women are perceived by jurors in the criminal justice system.

Within the criminal justice system, gendered schemas have interesting interactions with juror perception .The latest Bureau of Justice Statistics Report states that women comprise fifty-one percent of the population but only constitute less than twenty percent of convicted felony defendants (Greenfeld & Snell, 2000). This data supports the widely held belief, or gendered stereotype, that women do not commit as much crime as men. Consequently, juror perception of women defendants may employ a greater variation of contending heuristics in comparison to their perception of male defendants. Research data on juror perception of female defendants has been inconsistent and varies with type of crime, defense or circumstance, and defendant masculine qualities (Strub & McKimmie, 2016). By contrast, research on juror perception of male defendants has yielded stable results in prior years (Strub & McKimmie, 2016.) This could be indicative of competing heuristics that can alter perception depending on which heuristic is utilized at the time of making a specific judgment call.

Digging deeper into this phenomenon, researchers conducted an experiment to see if jurors would judge defendants more harshly if defendants deviated from their gendered stereotypes (Strub & McKimmie, 2016). Participants were given a hypothetical murder case to review and asked to decide on the verdict of the case. They were then asked to rate the extent to which they felt the evidence was convincing. The case transcripts were the same for all participants however researchers manipulated the sex (male or female) and description (masculine terms or feminine terms) of the defendants. Results showed that male defendants were overall more likely to be given a guilty verdict in comparison to women. Curiously, evidence was seen as more convincing when women were perceived to be deviating from their gendered stereotype (described in masculine terms). The findings suggest that men were judged using the heuristics for criminals while women were judged using the heuristic of gendered norms. These findings make it difficult to assess how participants will react to a gender manipulation in a study.

Receiving the answer key to a test could be described a college students’ dream regardless of gender, however, one must consider that there could be competing sets of heuristics employed in any study involving gendered norms. For example, a female Facebook user may be viewed as a damsel in distress in need of saving after she has worked so hard to end up in an impossible predicament, a supportive environment could prime the observer to employ this heuristic. In study one, supportive commentators made a significant consensus effect when compared to both the opposed and mixed conditions for the variable “Abigail was wrong”. According to research, it is possible that the Facebook user did not appear to be congruent with the initial thought of guilt because of her gender (Strub & McKimmie, 2016). Nonetheless, cheating on an exam could be viewed as deviation from a woman’s gendered norm and result in harsher judgement. Equally, men could automatically be associated with guilt because of their gender. But in situations less severe than murder, a perceiver may have re-categorized the situation with another heuristic (Strub & McKimmie, 2016). Many participants may remember the age-old heuristic of – boys will be boys – and not find the cheating behavior to be reproachable. These heuristics along with other variables could have an interesting interaction in study two.

This replication study with extension sought to replicate the results from study one while investigating whether the gender of the person experiencing the dilemma would alter how participants judge their actions. Again, participants were given a Facebook profile belonging to a student (Abigail/Adam) who mistakenly received the answer key to an exam, used it, and caused the other students to fail the test due to lack of a curve. The dilemma was again posted on Facebook seeking feedback on the best moral action to take. This time the opposed condition was removed leaving a 2 (comment condition: support v. mixed) X 2 (Facebook User Condition: Male v. Female) factorial design resulting in four conditions support/male condition (A), support/female condition (B), mixed/male condition (C), and mixed/female condition (D). Both Facebook pages were identical in nature apart from the user’s name and pictures. Comments also mirrored each other with only slight variations on specific words that could alter the sentence to show support or opposition.

It was predicted that the first main effect for comment condition would reveal participants in the mixed condition rating the Facebook user’s conduct as more wrong, more immoral, more unethical, more unacceptable, less understandable, less reasonable, and less appropriate than in the support condition, regardless of the Facebook users’ gender.

The second main effect prediction was that participants would see the cheating behavior of the female Facebook user as more wrong, more immoral, more unethical, more unacceptable, less understandable, less reasonable, and less appropriate than the male Facebook user.

The final hypothesis was on the interaction of Facebook user gender and comment condition. Researchers believed that participants would see the user as more wrong, unethical, immoral, etc. in the mixed/female condition (D) than all other conditions. The second-most “wrong etc.” was expected to be in the mixed/male condition (C), followed by the support/female condition (B). Finally, least “wrong etc.” would be participants in the support/male condition (A).

Methods Study Two

Participants

Study two consisted of 189 participants, of which 34% ( n = 65) were university students. Participant gender consisted of 40% ( n = 75) males and 58% ( n = 110) females. The ages ranged from 16 to a maximum of 70 with an average of 27.35 years ( SD = 10.13). The sample population consisted of 58% Latino/a ( n = 110), 15% Black ( n = 28), 19% White ( n = 36), 4% Asian American ( n = 7), 2% Others ( n =4), and 2% Indigenous ( n = 3). See Table 6 .

Table 6

Study Two Demographics

Materials and Procedure

People in this study were recruited to participate in an online study using Qualtrics. On the first page, participants were informed of the purpose, procedure, and risk associated with the survey. After consenting to take the survey, participants were taken to Part One, a mock Facebook page. The backstory of the Facebook dilemma was copied from study one. The study two Facebook pages used in each condition were also identical to study one with the exception of the independent variable manipulations. In this study, the Facebook user was either Abigail (female name and profile picture) or Adam (male name and profile picture) providing our first independent main variable. Our second main variable was a manipulation of comments left by the users’ friends. In this study, there were only two comment conditions (Mixed, Supportive), as opposed to three conditions in study one. Again, comments mirrored each other with only slight variations on specific words that could alter the sentence to show support or opposition. The first condition (support), revealed a consensus of eight comments which included praise for the Facebook users’ actions (don’t feel bad, you got lucky…) In the second condition (mixed) there were four supportive and four opposed comments with no clear consensus (don’t feel bad, you got lucky, you can’t take the grade, your integrity). These manipulations provided our 2 (comment condition: support v. mixed) X 2 (Facebook User Condition: Male v. Female) factorial design resulting in four conditions support/male condition (A), support/female condition (B), mixed/male condition (C), and mixed/female condition (D).

After reviewing one of the randomly assigned Facebook conditions, participants continued to Part Two, the first survey portion of the study, with instruction not to click back to the previous page to protect the integrity of the attention check questions. Part Two consisted of the same seven statements of judgment from study one (Their behavior was wrong, Their behavior was reasonable, etc) to be answered by a six-point Likert scale (1 = strongly disagree, 6 = strongly agree). The next page in the survey, Part Three, was also copied from Study One with the same Likert scale asking participants to rate how strongly they would agree with the same 5 statements of advice for the Facebook user (I would give them the same advice that their friends gave them, I would remain silent) and how strongly they would agree with the same 7 statements of their impression of the Facebook user (The Facebook owner seems sincere, the Facebook owner seems competent). In Part Four of the study, participants were asked the same demographic information as in study one.

Again, in this study there were over 20 dependent variables; however, many were ignored for this analysis. The main focus was on whether participants passed both manipulation checks, whether they agreed with the perceived consensus in their condition, and how they responded to two particular statements in Part Two (Their behavior was wrong, Their behavior was appropriate).

Results Study Two

The first manipulation check used the comment condition (Support vs. Mixed) as the independent variable and the feedback recall (Feedback supported the cheater vs. Feedback was mixed) as the dependent variable, the results were significant, X2(2) = 25.73, p < .001. Most participants who reviewed supportive comments were able to correctly choose the supportive feedback (64%) while most participants who reviewed mixed feedback were able to correctly choose mixed feedback (65%). However, Phi showed only a moderate effect. Some participants may not have been aware of the manipulation. See Table 7.

Table 7

Crosstabs and Chi Square – Study Two

The next crosstabulation was done using Facebook user gender as the independent variable and the gender manuplation response as the dependent variable, the results were also significant, X2(2) = 118.64, p < .001. Participants who received a Female Facebook User were able to correctly recall her gender (82%) while Participants who received a male Facebook user were able to correctly recall his gender (88%). Phi showed a strong effect suggesting participants were mostly aware of the manipulation as intended. See Table 8.

Table 8

Crosstabs and Chi Square – Study Two

For the first main analysis, researchers ran a 2 X 2 ANOVA with Facebook user gender and Comment condition as the independent variables and level of agreement with the statement “their behavior was wrong” as the dependent variable. There was a significant main effect for comment condition, F(1, 185) = 9.15, p = .003, with the supportive comments ( M = 3.56, SD = 1.40) differing from the mixed comments ( M = 4.13, SD = 1.21). This reveals that participants more strongly agreed that the behavior was more wrong when the consensus was mixed than when comments were only supportive. There was also a significant Facebook user gender main effect, F(1, 185) = 6.61, p = .011, with the male Facebook user ( M = 4.09, SD = 1.00) differing from the female Facebook user ( M = 3.61, SD = 1.56). This shows that participants more strongly agreed that the behavior was wrong when the Facebook user was male than when the user was female. Unfortunately, there was no significant interaction between Facebook user gender and comment condition for this variable, F(1, 185) = 2.03, p = .156. This shows that means did not differ between participants in the female supportive condition ( M = 3.19, SD = 1.54), female mixed condition ( M = 4.02, SD = 1.48), male supportive condition ( M = 3.94, SD = 1.13), and male mixed condition ( M = 4.23, SD = 0.84). See Table 9.

Table 9

ANOVA “Their behavior was wrong”- Study Two

The final analysis was a 2 X 2 ANOVA with Facebook user gender and Comment condition as the independent variable and level of agreement with the statement “their behavior was appropriate” as the dependent variable. There was a significant main effect for comment condition, F(1, 185) = 7.82, p = .006, with the supportive comments condition ( M = 4.40, SD = 1.03) showing a higher level of agreement with the statement than the mixed comments condition ( M = 4.06, SD = 0.68). There was also a significant Facebook user gender main effect, F(1, 185) = 12.02, p < .001, with participants rating this behavior more appropriate in the female Facebook user ( M = 4.44, SD = 0.85) than in the male Facebook user ( M =4.02, SD = 0.88).

The main effect was qualified by a significant Cheater Gender X Comment interaction, F(1, 185) = 4.41, p = .037. First, simple effects showed that support participants rated the cheater’s behavior as more appropriate in the female user condition ( M = 4.74, SD = 0.74) than support participants in the male user condition ( M = 4.06, SD = 1.17), F(1, 92) = 11.42, p = .001. Second, simple effects showed that mixed participants did not differ in their ratings of whether the cheater’s behavior was appropriate in the female user condition ( M = 4.15, SD = 0.85) and male user condition ( M = 3.98, SD = 0.44), F(1, 93) = 1.44, p = .234. Third, simple effect tests showed that participants exposed to the male profile, did not differ in their ratings of whether behavior was appropriate in the support ( M = 4.06, SD = 1.17) and mixed condition ( M = 3.98, SD = 0.44), F(1, 92) = 0.22, p = .642. Finally, for participants in the female user condition, simple effect tests showed that participants felt the cheater’s behavior was more appropriate in the support condition ( M = 4.74, SD = 0.74) than participants in mixed condition ( M = 4.15, SD = 0.85), F(1, 93) = 13.44, p < .001. Overall, data shows that participants rated the moral behavior as more appropriate for females in the support condition than in the mixed condition; a difference not seen in the male data. For mixed conditions there was no difference between males and females. See Table 10.

Table 10

ANOVA “Their behavior was appropriate” - Study Two

Discussion Study Two

The comment manipulation was strong enough to support predictions by moderately swaying participant responses; mixed condition participants rated the Facebook user’s conduct as more “wrong” and “less appropriate” than they did in the support condition. However, no evidence was found to support the hypothesis that the behavior of female Facebook users would be rated as more “wrong” and “less appropriate” than males. While the analysis of the manipulation check made it clear that the gender manipulation was noticed and correctly recalled, the effect was opposite to what was predicted. Males were rated as significantly more “wrong” and less “appropriate” than females; consequently, results did not support research predictions for the interaction of variables in this analysis. There as no interaction for the “wrong” variable but there was a significant interaction for the “appropriate” variable. Interestingly, the difference between gender conditions were not noticeable in mixed conditions for the “appropriate” variable. This finding suggest that the consensus had the strongest effect on participant feedback when comments were unanimously supportive of the female user.

General Discussion

In study one, researchers correctly predicted that participants would notice the studies manipulation in all three conditions and that their feedback would mirror their respective conditions. However, results also revealed that when participants were asked if Abigail’s behavior was wrong the support condition differed from the mixed and opposed conditions; yet the opposed and mixed conditions did not significantly differ from each other. This result was unexpected but can be explained if the level of consensus was indistinguishable between opposed and mixed conditions. In previous replications of the ash line experiment, reported levels of conformity directly corelated with the number of confederates placed in the room (Asch, 1951, as cited by Kundu & Cummings, 2013). With one or two confederates the conformity effect was minimal but with 3 or four confederates the effect grew. It is possible that the opposed condition was not worded clearly enough to prime participants with what they perceived to be a strong consensus.

Study two copied the mixed/ support comments from study one verbatim. The 2 X 2 ANOVA in study two revealed that the support/mixed condition manipulation check only showed a moderate effect. Again, this could have altered the strength of other variables and interactions. Previous research shows that negative comments could have a greater effect than positive comments (Winter et al., 2015). Study two may have seen a greater condition effect if only the support and opposed conditions were, and if the opposed consensus appeared stronger to participants.

It was predicted for both studies that participants in all conditions would agree to give Abigail the same advice that her friends gave to her. Results support this hypothesis; however, no condition resulted in a mean close to the strongly agree score on the Likert scale. Instead, means on both studies for all conditions scored closely to the statements “I somewhat disagree” or “I somewhat agree”. Standard deviations were also somewhat comparable in all conditions apart from the Female condition in study two being slightly higher and the standard deviation in study one (all female) being slightly lower. The polar like standard deviations found in female conditions reflects previous research done with jurors not showing stable patterns of judgment when it comes to female defendants, suggesting that different heuristics can also come into play when making a moral judgment (Strub & McKimmie, 2016). In study two, males were rated as significantly more “wrong” and less “appropriate” than females; the effect was opposite to what was predicted; however, it supported prior research showing that men are more steadily associated with guilt than women (Strub & McKimmie, 2016). Despite the heuristic used, or gender judged, results were still in accordance with researcher predictions about comment conditions. Participants in both studies rated the Facebook user’s conduct as slightly less “wrong” and more “appropriate” in support conditions than in opposed or mixed conditions showing that participants moderately conformed to the perceived consensus of their group. In study two, there was no interaction effect for the “wrong” variable but there was a significant interaction for the “appropriate” variable with the strongest effect seen when participants were primed with supportive comments for the female user. One could speculate that the supportive comments primed participants to employ a heuristic which painted the female user as someone in need of help instead of chastisement.

In further exploration of the limitations in this study, it is possible that the statements used to distinguish the mixed and opposed conditions in study were not very different from each other. The comments manipulated to show consensus were long and could have provoked bias or ambiguity for readers. For example, many statements in the oppositional and mixed conditions began with phrases that could be perceived as supportive but lead into opposition (You know that if you didn't get the answer key another student would have; and then they would have the highest grade and you'd lose out on the curve, how would you feel then? , Listen, it's not like you intended to cheat on the exam the professor should have checked to make sure he was only handing out blank exams. His mistake! But your integrity, don't take the grade!). Though the statements were overall unsupportive, the conclusion about the statement was made after being primed with supportive material. This effect could have carried over to the supportive condition with two specific comments that appear ambiguous or oppositional (You know that if you didn't get the answer key another student would have, and then they would have the highest grade and lose out on the curve, don't feel too bad; If it were me I'd tell the prof…. NOT! don't be crazy you might blow the next exam so it will all even out in the end.) Another limitation lies between the context, location, and demographics of the study. It is conceivable that college students may be hesitant in being truthful on campus or online due to strict college guidelines against cheating behavior. Similarly, a student in a difficult class may have been emotionally primed by his or her own wishful thinking. Perhaps a future study can be replicated with shorter and stronger consensus statements along with a different moral dilemma, such as watching a passerby drop their wallet or refusing to change seats on a plane to allow a mother to sit close to her fearful child.

References

Cohen, J.B., & Golden, E. (1972). Informational social influence and product evaluation . Journal of Applied Psychology, 56(1), 54–59. https://doi-org.ezproxy.fiu.edu/10.1037/h0032139

Greenfeld, L. A., & Snell, T. L. (2000). Women offenders. Bureau of Justice Statistics.

Hingston, S. T., & Noseworthy, T. J. (2018). Why consumers don't see the benefits of genetically modified foods, and what marketers can do about it.  Journal of Marketing, 82(5), 125-140. https://doi-org.ezproxy.fiu.edu/10.1509/jm.17.0100

Kelly, M., Ngo, L., Chituc, V., Huettel, S., & Sinnott-Armstrong, W. (2017) Moral conformity in online interactions: rational justifications increase influence of peer opinions on moral judgments, Social Influence, 12:2-3, 57-68, https://doi-10.1080/15534510.2017.1323007

Kundu, P., & Cummins, D.D. (2013) Morality and conformity: The Asch paradigm applied to moral decisions, Social Influence, 8:4, 268-279. https://doi-10.1080/15534510.2012.727767

Signorella, M. L., & Frieze, I. H. (2008). Interrelations of gender schemas in children and adolescents: attitudes, preferences, and self-perceptions. Social Behavior & Personality: An International Journal, 36(7), 941–954. https://doi-org.ezproxy.fiu.edu/10.2224/sbp.2008.36.7.941