Introduction Summaries The introduction and evolution

Introduction Summaries The introduction and evolution

The introduction and evolution of information and communication technologies

(ICTs) have delivered a plethora of benefits such as the increase of communication and

innovative means of marketing products and services. The impact of information systems

(IS) on everyday human life is considered a utility similar to that of water, electricity, gas

and telephone (Buyya, Yeo, Venugopal, Broberg, & Brandic, 2008). Yet, despite the

advantages provided by information systems, there are disadvantages that are potentially

dangerous to the individuals that use them. For instance, information privacy is a major

the personal information of their customers to profile them, whereby organizations could

develop stronger rapport with their customers, thereby growing their business (Awad &

Krishnan, 2006; Culnan & Armstrong, 1999). Yet, the trade of personal information is

considered a “double-edged sword” whereby personal information could be both an asset

to an individual and a risk (Malhotra, Kim, & Agarwal, 2004). Organizations could use the

personal information they have collected about an individual unethically such as selling it

to third parties (Culnan, 1993; Smith, Milberg, & Burke, 1996), as well as reuse the

collected personal information legally, but in ways that are unsanctioned by the owners

(i.e. clients/customers) of the personal information (Culnan & Williams, 2009). Due to

these risks, consumer reports and surveys have often revealed that individuals have great

concerns over the privacy of their personal information (Dinev & Hart, 2006;

indicating that the adoption and growth of ecommerce have not reached its full potential

(US Census Bureau News, 2015). Similarly, while governments have invested in ehealth

for safer and more efficient healthcare systems (Angst & Agarwal,

2009), patients are concerned about their personal medical records (Bishop, Holmes, and

Kelley, 2005). It is therefore necessary to understand how both individuals and organizations

could capitalize on the benefits provided by information systems, without compromise to

personal information.

Researchers identified the privacy paradox as a fundamental issue whereby a

conflict of interest exist between individuals’ stated intentions and actual behavior

regarding privacy related decisions (Norberg, Horne, & Horne, 2007). The privacy

information privacy. The privacy paradox becomes a key issue in understanding why

individuals are willing to disclose their personal information, and if their decisions are

1.2. Research Problem and Argument

This study investigated the privacy paradox to better understand individuals’ decisions

to withhold or disclose their personal information. Researchers have studied the privacy

paradox in the context of ecommerce (Acquisti, 2004; Dinev & Hart, 2006;

Norberg et al., 2007), personalization (Awad & Krishnan, 2006; Culnan & Armstrong,

1999), ehealth (Anderson & Agarwal, 2011; Angst & Agarwal, 2009), location based

services (Xu, Teo, Tan, & Agarwal, 2012), and web 2.0 technologies (Dinev, Smith, &

Xu, 2009). Yet, despite the impact the context may have on influencing individuals’

decisions to disclose personal information, privacy concerns are consistent in negatively

individuals are rational decision-makers (Acquisti, 2004; Acquisti & Grossklags, 2005), where

multiple factors may subvert the negative effect privacy concerns have over personal

information disclosure (Smith et al., 2011). Trust has often been found to be a plausible

explanation as to why individuals would disclose their personal information despite claims of

concerns (Belanger, Hiller, & Smith, 2002; Dinev & Hart, 2006; Pavlou, Liang, & Xue, 2007;

Van Slyke, Shim, Johnson, & Jiang, 2006). Essentially, an individual’s propensity to trust an

override their privacy concerns. Yet, trust and privacy concerns are not the only factors to be

considered since privacy risk has also been observed to be a salient factor in the privacy

paradox (Dinev & Hart, 2006; Malhotra et al., 2004; Norberg et al., 2007). While privacy

concerns and privacy risks are closely related, Dinev and Hart (2006) considered them to be

distinct. Furthermore, risk may not be one dimensional as often assumed, but may involve

both the assessment of loss as well as the probability of avoiding a risky action (Peter &

individuals’ decisions to disclose their personal information consists of contrary but salient

factors, which included trust, privacy concerns, and privacy risks. Additionally, Dinev and

Hart (2006) considered personal interest as a factor that could override the negative impacts of

privacy risk and privacy concerns. Similarly, Van Slyke et al. (2006) found similar factors of

trust, privacy risk, and privacy concerns as key indicators to individuals’ decisions to disclose

personal information. Yet, the cost-benefit assumption is limited by the reality that humans are

not purely rational decision-makers (Acquisti & Grossklags, 2005).

Acquisti (2004) explained that several psychological deviations limit rationality

from individuals. Acquisti (2004) explained that people are subject to bounded rationality,

decision, and even if they did, they would not be able to accurately process all of these

parameters. Moreover, cognitive biases such as hyperbolic discounting, where individuals

prefer short-term gratifications, affect rationality in privacy-related decisions (Acquisti,

2004). Evidently, research from the cognitive neuroscience field has identified a number

of brain areas that are correlated with mental processes such as perceptions of risk and

trust (Dimoka, Pavlou, & Davis, 2007; 2011). Furthermore, these brain activations occur

As such, this research argued that individuals’ disclose their personal information based

on their cognitive disposition, which includes rational and emotional mental processes.

1.3. Importance of Research Problem

1986). Specifically, the evolution of ICTs allows for increased surveillance, computation,

storage and retrieval of individuals’ personal information (Mason, 1986). Organizations

individuals’ personal information (Li & Sarker, 2006; Mason, 1986). Additionally, the

growth of the internet provides convenient business transactions for individuals, as well

as increased communications. Moreover, individuals use the internet and other innovative

web applications to become content providers (Hong & Thong, 2013). Essentially, this

leaves individuals vulnerable to privacy-related threats (Culnan & Williams, 2009; Hong

& Thong, 2013). While individuals continue to voice concerns for the privacy of their

2011).

their personal information, yet continuing to disclose their personal information for

certain benefits, termed the privacy paradox, has been extensively studied by multiple

Belanger & Crossler, 2011; Dinev & Hart, 2006, Norberg et al., 2007; Smith et al., 2011).

It was found that individuals’ privacy beliefs and their associated privacy-related behavior

could be classified into three categories: privacy fundamentalists, privacy pragmatists,

and privacy unconcerned (Harris Interactive & Westin, 2002). While privacy

fundamentalists were more skeptical of disclosing their personal information, and fought

costs in disclosing their personal information (Angst & Agarwal, 2009; Awad &

Krishnan, 2006; Harris Interactive & Westin, 2002). Individuals who were identified as

privacy unconcerned, however, would readily disclose their personal information, without

much concern or emphasis to the concept of information privacy (Angst &

Agarwal, 2009).

The beliefs of individuals with regards to information privacy, and subsequently

are privacy concerns, privacy risk, institutional trust (i.e. the propensity to trust), and

uncertainty (Awad & Krishnan, 2006; Belanger et al., 2002; Dinev & Hart, 2006;

Malhotra et al., 2004; Norberg et al., 2007; Pavlou et al., 2007; Van Slyke et al., 2006).

Additionally, studies have indicated that other factors such as culture, regulations,

organizations’ privacy practices (such as privacy seals, privacy statements, and organizations’

transparency of the use of collected personal information), and methods of persuasion were

related to privacy beliefs (Awad & Krishnan, 2006; Angst & Agarwal,

2009; Dinev, Bellotto, Hart, Russo, Serra, & Collautti, 2006; Li et al., 2006; LaRose &

While an individuals’ privacy beliefs, along with other related antecedents may

inform privacy-related behavior, some researchers point out that privacy-related

Grossklags, 2005). As argued by Acquisti (2004), multiple cognitive biases, such as

cognitive overload (inability to cognitively process all parameters of a given situation),

may affect individuals’ privacy-related decisions. Specifically, individuals’ decisions are

not purely rational, which may often result in outcomes that seem contrary to their beliefs.

Similarly,

influence such beliefs could provide an explanation to the privacy paradox, a better

understanding of individuals’ privacy-related decision-making could be elicited by

examining individuals’ perceptions in privacy situations. Specifically, while intrinsic and

extrinsic factors may be antecedents of privacy concerns and privacy-related decisions, the

momentary cognitive and affective state before an individual makes a decision with

regards to information privacy should be considered. Current literature has investigated

individuals’ privacy beliefs (Dinev & Hart, 2006; Van Slyke et al., 2006), the factors

which may affect these beliefs (Angst & Agarwal, 2009; Li et al., 2011), as well as the

cognitive limits of individuals in privacy-related situations (Acquisti, 2004; Anderson &

Agarwal, 2009). However, a gap in the literature exists in observing how individuals’

perceptions are formed and relate to one another when they are in a situation requiring

indicated that human behavior is often influenced by the processing of some external

stimuli, before conscious thought (Dimoka, 2010; Dimoka et al., 2007; 2011; Sur &

Sinha, 2009; Vance, Eargle, Anderson, & Kirwan, 2014). Essentially, while long held

beliefs and other intrinsic and extrinsic factors may play a role in privacy-related

decisions, the state of an individual produced by a privacy-related decision (such as

whether to disclose personal information to an online seller or for entry into an electronic

With the proliferation of data produced at an exponential rate, information privacy

invest in and promote the use of various ICTs for data mining purposes to remain

competitive (Awad & Krishnan, 2006; Culnan & Armstrong, 1999). So as to not waste

based on knowledge derived from understanding the privacy paradox, to motivate

privacy as a societal issue within the information age led to the development of the

federal trade commission (FTC) fair information practices (FIPs) which described four

for the protection of individuals’ personal information (Liu & Arnett, 2002). While the US

FIPs allowed for organizations to regulate themselves by establishing the FIPs within

their business processes, other regulations and laws were established for different industry

sectors to better protect the personal information organizations collected (Culnan &

Williams, 2009). For example, the Health Insurance Portability and Accountability Act

In fact, the European Union (EU) differs from the US in the view of privacy, where they

focus on the implementation and enforcement of safeguards for organizations to fairly

manage individuals' personal information (Rose, 2006). Also, despite the view that

individuals' privacy is a human right, when examining consumer behavior, a privacy

paradox was observed whereby individuals continue to trade their personal information for

specific benefits (Smith et al., 2011). This lead to the shift in the definition that the value of

privacy was more regarded as a commodity than a human right (Smith et al., 2011).

Armstrong, 1999; Dinev & Hart, 2006; Norberg et al., 2007). Essentially, the

commoditybased definition holds that an individual's personal information is valuable, but

value. Assigning an economic value to information privacy differentiates the definition of

privacy as a commodity from privacy as a right (Smith et al., 2011). Specifically, while

both definitions regard information privacy as an objective value in society, when privacy

is defined as a commodity, it essentially means that the individual may lose ownership of

his/her personal information. Conversely, privacy as a human right retains the notion that

even if an individual were to disclose his/her personal information to some other person or

personal information, which is influenced by situations which includes self-ego,

environmental, and interpersonal factors (Laufer & Wolfe, 1977). With regards to

information privacy, a state-based definition posits that individuals’ subjective cognitions

of the situations in which they are asked to disclose their personal information would

A widely accepted definition of privacy within the field of information systems, as

well as other fields such as philosophy, social and political sciences, psychology and

his/her personal information (Smith et al., 2011). Specifically, as long as an individual

assumes that he/she can control the use of his/her personal information, and minimize the

risks posed by disclosing it, he/she would be willing to provide his/her personal

information to other entities. The control-based definition of general privacy is rooted in

the definition by Westin (1967), as well as that of Altman (1975) as "the selective control

of access to the self" (p. 24). Margulis (1977) captured the essence of the control-based

Laufer and Wolfe (1977), an individual may assert some level of control over his/her

personal information, however that does not equate to privacy. Control, however, may be

a key factor in achieving privacy (Laufer & Wolfe, 1977; Smith et al., 2011).

2.2.3. Privacy within the Information Systems Field

definition to information privacy allows researchers to better understand how the concept

of information privacy relates to other factors (Pavlou, 2011; Smith et al., 2011).

is by understanding its meaning and relationship with other factors that drive individuals'

behavior towards specific tasks.

adopted a right-based definition (Liu, Marchewka, Lu, & Yu, 2005; Malhotra et al.,

adopted a control-based definition of information privacy assuming that individuals’

information disclosure (Belanger & Crossler, 2011; Culnan & Bies, 2003; Dinev & Hart,

2004; Van Slyke et al. 2006). Alternatively, a growing number of studies adhere to the

explains the current paradox of individuals’ personal information disclosure, yet high

conflicting findings, the concept is still inconsistent and requires much study (Pavlou,

2011; Smith et al., 2011). Researchers have often used privacy concerns as a means of

and near impossible to measure by itself (Dinev et al., 2009; Smith et al., 2011). While

Dinev et al. (2009) regarded privacy concerns as inadequate to fully understand

information privacy, privacy concerns as a construct is a representation of the negative

perceptions an individual has over the collection and use of his/her personal information

he/she may be required to disclose to another person or organization.

have treated information privacy as a negative factor (Dinev et al., 2009). This is further

justified based on the evidence presented from polls whereby individuals claim a high level of

concern for the privacy of their personal information (Dinev & Hart, 2006; Madden et al.,

2007; Smith et al., 2011). Similarly, numerous studies have found that privacy concerns inhibit

individuals from disclosing their personal information (Dinev & Hart, 2006; Li et al., 2011;

Pavlou et al., 2007; Van Slyke et al., 2006).

of concerns individuals have over their personal information: collection, errors, secondary

use, and improper access (Smith et al., 1996). Collection refers to the extensive collection

the use of such personal information of individuals for purposes apart from that which

was stated (Smith et al., 1996). This would include selling the personal information to

third party organizations. Individuals are also concerned about the access others may have

to their personal information, despite the lack of proper authorization. Finally, errors refer

to the concern individuals may have over the deliberate and accidental errors to their

personal information, and the lack of safeguards towards such cases (Smith et al., 1996).

modeling the CFIP as a higher-order construct would allow for more accurate findings

withhold their personal information (Stewart & Segars, 2002). However, Malhotra et al.

(2004) explained that there were limitations to CFIP which did not account for the overall

personal information.

Malhotra et al. (2004) developed the Internet Users' Information Privacy

global information privacy concern (GIPC) and CFIP. Malhotra et al. (2004) explained

that GIPC suffered from the lack of dimensionality in explaining individuals' privacy

because of the internet could impact the dimensions of privacy concerns individuals may

have (Malhotra et al., 2004). As such, the IUIPC accounted for these limitations by

developing a better scale for measuring and studying individuals' privacy concerns.

collected personal information, IUIPC concentrated on the perceptions of fairness

Hart, 2006; Pavlou et al., 2007; Van Slyke et al., 2006), privacy concerns may be lacking to

not represent individuals' overall perception of information privacy since a person's privacy

may not be violated despite his/her high levels of privacy concerns (Dinev et al., 2009).

privacy to any individual or society (Dinev et al., 2009). Essentially, the overall concept of

privacy attitudes were influenced by privacy beliefs which were preceded by privacy

values. Similarly, Belanger and Crossler (2011) indicated that individuals' privacy attitudes

saliency of privacy concerns with regards to information privacy, there are other measurable

factors of information privacy that might provide a better understanding of the concept.

thus contributing to organizational growth (Awad & Krishnan, 2006; Culnan &

Armstrong, 1999). Similarly, from the individual and societal perspective, there are

(Angst & Agarwal, 2009). It is only by understanding the privacy paradox, organizations,

government, and society could maximize the benefits of information exchange and

minimize the costs.

researchers would examine the privacy paradox with regards to online transactions

organizations and individuals, it can incur heavy privacy concerns (Awad & Krishnan,

the sensitivity of information is increased, there exists reluctance to personal information

disclosure, such as in the case of medical information (Anderson & Agarwal, 2011; Angst

context, privacy concerns have been found to have a negative effect in the disclosure of

The inconsistency posed by the privacy paradox has driven researchers to seek an

understanding as to why individuals would provide their personal information despite their

Norberg et al., 2007; Pavlou et al., 2007). Studies have indicated that as long as

individuals retain a level of control of their personal information, they will be willing to

disclose their personal information online (Awad & Krishnan, 2006; Culnan, 1993; 2000;

Culnan & Bies, 2003; Dinev & Hart, 2004). Culnan (1993) found that individuals with a

positive attitude towards secondary information use were different from those with

negative attitudes because of the level of control they had over their personal information.

personal information (Culnan & Armstrong, 1999; Lin & Wu, 2008; Malhotra et al.,

services and personalized advertisement, individuals were more willing to participate in

the former since it was more valuable, which essentially indicated that there was some

were carried out by researchers to elicit a more thorough understanding of individuals'

(2011) corroborated the arguments that individuals’ privacy behavior was not purely

rational.

privacy calculus with regards to the privacy paradox by including emotions as a

necessary factor for consideration. Anderson and Agarwal (2011) argued that the

more granular than has been explored in past privacy studies” (p. 471). As such,

privacy calculus; essentially, a major contribution from their study was that

being purely rational. Despite the argument of Anderson and Agarwal (2009) that

privacy within the healthcare context needed to be studied differently, their findings

when examining individuals’ privacy behaviors, despite the context.

2.5. Summary

Information privacy is a subset of the general field of privacy, which has been

defined by researchers as either a value or cognate-based (Smith et al., 2011).

individuals who claim concerns for information privacy, but continue to disclose their

2005). Studies have therefore integrated emotions as key factors in individuals’ privacy-

related decision-making (Anderson & Agarwal, 2011; Li et al., 2011). Essentially, a better

understanding of the privacy paradox could be achieved by considering both the rationality

and emotional thought processes behind individuals’ privacy-related decisions.

Chapter 3

Theory

3.1. Introduction

This chapter consists of the theory used for the development of the research

model and hypotheses for this study. Section 3.2 presents the theory used for this study,

how it was expanded upon using cognitive neuroscience, as well as relevant information

of the neuroanatomy of the human brain to better understand how cognitive neuroscience

is used to better explain the privacy paradox. Section 3.3 presents an overarching

research model for the study, along with the hypotheses, and their separate models.

The privacy calculus has often been used to explain individuals' privacy-related

decisions, particularly in understanding the privacy paradox, positing that individuals'

would trade their personal information based on a cost-benefit analysis (Culnan & Bies,

2003; Dinev & Hart, 2004; Laufer & Wolfe, 1977, Smith et al., 2011). The principles of

the privacy calculus have often been adopted by researchers greatly in extant literature in

the information privacy field (Bansal et al., 2010; Dinev & Hart, 2004; 2006; Dinev et al.,

2006; Li et al., 2011; Malhotra et al., 2004; Xu et al., 2010). Essentially, researchers have

often explained the privacy paradox using the privacy calculus, highlighting the numerous

The privacy calculus is limited in assuming individuals are purely rational decision

makers, yet, researchers have recognized the need to include emotions in investigating

Following the studies that sought to gain a better understanding of the privacy paradox,

this study adopted the extended privacy calculus model developed by Dinev and Hart

(2004) as the theoretical basis. However, the extended privacy calculus model is limited in

the assumption that individuals are purely rational decision-makers. To address this

limitation, the findings from cognitive neuroscience were applied to the extended privacy

calculus model. The introduction of cognitive neuroscience thus enhanced the explanation

privacy concerns, but influenced by trust and personal interest (Dinev & Hart, 2006).

Privacy risk was defined as the risk of opportunistic behavior towards the personal

information an individual disclosed, which was associated with the possibility of loss and

concern, but negatively with trust (Dinev & Hart, 2006). Essentially, if an individual

perceived more trust in disclosing his/her personal information, his/her perception of

privacy risk would be lower. Trust was one of the confidence and enticement beliefs,

which positively influenced individuals’ intentions to disclose personal information.

Personal interest is the other confidence and enticement belief factor that was defined as

the intrinsic motivation toward performing an action, and the satisfaction derived from

Figure 1. Extended Privacy Calculus Model developed by Dinev & Hart (2006)

3.2.2. Using Insights from Cognitive Neuroscience to Enhance the Extended Privacy

Calculus

explained the cognitive bias of hyperbolic discounting as a cause for individuals’

decisions to disclose personal information despite claiming concerns for their information

choice of long-term benefits (Acquisti, 2004). However, explaining the privacy paradox

by arguing for the saliency of cognitive biases, such as hyperbolic discounting, ignores

the rational aspects involved in privacy-related decisions. While individuals’ disclosure of

personal information may not be purely rational, findings from cognitive neuroscience

have explained that decisions are complex enough to involve multiple mental processes

(Dimoka et al., 2007).

emotional. Furthermore, Dimoka (2012) explained that individuals’ brain activity does not

relationship between brain activity and mental processes (Dimoka, 2012). For instance,

when an individual perceives some activity will yield specific rewards, several key brain

amygdala, nucleus accumbens, and caudate nucleus (Dimoka et al., 2007). The brain areas

privacy paradox, it would be essential to observe how individuals decide to withhold or

disclose their personal information.

individuals’ brain activity, and associate brain functions with mental processes. Popular

within brain mapping literature, is the use of the functional magnetic resonance imaging

have used the electroencephalogram (EEG) to track individuals’ electrophysiological

Table 1. Frequency bands of electric potentials

aves

Theta waves 4-8 Connected to feelings of deep and raw emotions. High

Beta waves 13-21 Relates to conscious thought, logical thinking, focus and

High Beta

waves

Gamma

waves

factors involved in the extended privacy calculus enhances the investigation of

individuals’ privacy-related decisions. Furthermore, other factors such as uncertainty and

distrust are important in considering since uncertainty was found to be an antecedent of

personal information disclosure (Pavlou et al., 2007), and distrust as a distinct factor from

trust (Dimoka, 2010). Thus, the findings of cognitive neuroscience were applied to the

extended privacy calculus model, to elicit a better explanation of the privacy paradox,

identifying both the involvement of emotions and rationality in individuals’ privacyrelated

decisions.

3.2.3. Neuroanatomy – Structure and Function of Key Brain Areas

telencephalon and diencephalon. The telencephalon is composed of the cerebrum,

cerebral cortex, and subcortical structures such as the basal ganglia (Cannon, 2012;

frontal lobe is the anterior part of the brain, the temporal lobe is located to the side, and

the parietal lobe is behind the frontal lobe, but above the occipital lobe. The occipital

lobe, which is located at the base of the brain, is primarily associated with vision (Demos,

Further division of the cerebral cortex into specific areas was developed by

Korbinian Brodmann in 1908 into a cytoarchitectural map consisting of Brodmann Areas

(BA) labeled 1 to 52 (Cannon, 2012). The BA’s were developed based on the organization

of neurons, and have since been the most widely accepted cytoarchitectural map of the

cerebral cortex, which, over the years, have undergone re-evaluations and further

divisions (Cannon, 2012). BA’s are responsible for many different functions ranging from

Localization of brain activity may not be the same across neuroimaging technologies;

however, past research can provide an idea as to where neural activity may take place in

specific situations. Thus, these hypotheses are developed more as a guideline to where

brain activity might occur. Specifically, exploring neural activity for specific mental

processes might produce varied results from past literature based on a number of factors

such as the design of experiments and neuroimaging tools selected. As explained by

Cannon (2012), the brain is a complex system of systems, whereby findings in current

neuroscience literature can be challenged as new findings can refute previous

Researchers in the field of information privacy have often treated privacy risk as a

one-dimensional construct (Dinev & Hart, 2006; Malhotra et al., 2004; Norberg et al.,

2007; Smith et al., 2011). However, privacy risk may actually be a multidimensional

construct whereby an individual's perception of risk may assess the likelihood and

severity of consequences in engaging or avoiding a risky action (Peter & Tarpey, 1975;

Smith et al., 2011). Privacy risk is essentially an individual's perception of risk but in the

context of information privacy. According to Dinev and Hart (2006), privacy risk wis

defined as the possibility of loss, which is related to the uncertainty caused by the

H1b: Privacy Risk is associated with brain activity in the insula cortex. H1c:

which are different from the neural correlates in risky situations. Pfeffer and Salancik

(1978) explained uncertainty as the inability to accurately anticipate the future state of a

situation because of a lack of enough information. Specifically, when using ICTs, such as

ecommerce, uncertainty may exist when an individual is unsure of the outcome if he/she

were to disclose his/her personal information. This is often due to a lack of information

regarding the parameters of such a situation, such as the case with buyers and sellers in

ecommerce transactions (Acquisti, 2004; Pavlou et al., 2007). Yet, uncertainty is different

from ambiguity and risk. An individual may be privy to a great deal of information when

cognitive processes of high activations in the ventromedial prefrontal cortex and inferior

parietal lobule (right BA 40), but low activations in the dorsolateral prefrontal cortex

(Deppe et al., 2005).

Furthermore, findings in neuroscience suggest that risky situations involve brain areas

that are involved in strong negative emotions (such as the insula cortex) as well as reward

centers (Knuston et al., 2001; Paulus & Frank, 2003). While the full extent of the role

privacy risk may be unknown in relation to privacy-related situations, the following

hypothesis was developed based on findings from prior literature.

H7b: The neural correlates of trust are positively related to personal information

influence individuals to withhold their personal information.

H9: The neural correlates of personal interest are positively related to personal

H10a: The neural correlates of uncertainty are positively related to privacy

H10b: The neural correlates of uncertainty are negatively related to personal

Privacy concerns are often used in information privacy field as an appropriate

measure of information privacy (Smith et al., 2011). Dinev & Hart (2006) defined privacy

concerns as the concerns an individual has over the opportunistic behavior of another

entity to which an individual would disclose his/her personal information. Dinev and Hart

the collection, unauthorized secondary use, improper access and errors of personal

dimensions of privacy concerns in the era of the internet were based on the concerns for

the collection and control of personal information, as well as the individual's awareness of

how the personal information he/she has disclosed to an entity is handled. It is assumed

that the greater an individual's degree of privacy concerns, the more likely he/she would

withhold his/her personal information. Figure 6 depicts the research model that was tested

in the study (except for hypothesis 12).

Based on the findings of cognitive neuroscience, the constructs or mental

perceived by an individual. In the same manner that both distrust and trust can be

varying degrees. However, the effect of an independent variable on the dependent

are referred to as interaction effects. It should be noted that the following hypotheses on

the dependent variables. For instance, the interaction effects would not alter hypotheses

such as 6b, which posits that risk would negatively influence personal information

disclosure, but would determine the amount of change in the dependent variable based on

the interactions between independent variables.

H12a: An interaction effect would exist between the independent variables

H12b: An interaction effect would exist between the independent variables (privacy

3.4. Summary

The extended privacy calculus model developed by Dinev and Hart (2006) was

used as the theoretical basis for this study. However, the extended privacy calculus model

is limited in assuming individuals are purely rational decision-makers. To address this

limitation, the findings of cognitive neuroscience were added to the extended privacy

calculus. The findings of cognitive neuroscience aid in identifying the neural correlates of

mental processes, which predict individuals’ privacy-related decisions. Identifying these

neural correlates allow for examining the nature of these mental processes (such as

subjects were allowed to take a break if they felt uncomfortable or stressed, at which

point the emotiv EPOC+ was removed and recordings were stopped. When subjects were

ready to resume, the emotiv EPOC+ was once again placed on the scalp, and all

electrodes tested for data acquisition. In total, there were only two subjects who took a

break. If at any point an electrode stopped collecting data or became faulty, the subjects

were asked to stop and the electrode(s) was re-adjusted until data was once more

captured. At times, this entailed an electrode moving out of position, bad connectivity

using Bluetooth, or the electrode becoming dry.

Taken as one experiment, the least amount of time taken by one subject was

and when they made their decisions were recorded for the creation of epochs for later

4.4.1. Experiment 1: Neural Correlates

utilized review profiles of simulated organizations and/or websites. Review profiles have

been used in prior studies to elicit specific perceptions and observe subjects’

decisionmaking (Ba & Pavlou, 2002; Dimoka, 2010). For example, Dimoka (2010) used

feedback profiles to examine subjects’ neural activity to the constructs of trust and

distrust. In the context of ecommerce transactions, feedback profiles were found to be

adequate treatments since they help buyers to accept or reject sellers (Pavlou & Dimoka,

2006). Similarly, the internet has been used by reviewers to provide useful feedback and

reviews for organization and websites.

Conditions 1 to 3 reflected the factors of privacy risk, trust and distrust,

respectively. Essentially, the profiles for conditions 1 and 3 consisted of negative reviews

the simulated organization or website. As explained above, uncertainty is related to

Condition 5, personal interest, was measured by allowing subjects to choose either an

ecommerce product category that they would be highly interested in obtaining from

variety of reasons such as convenience or the desire to obtain a specific item that can only

be found online, an accurate means of measuring personal interest would be to actually

let subjects choose something in which they have a vested interest. The profiles for

experiment 1 are located in Appendix B.

For each profile (conditions 1-4), subjects were exposed to one review comment,

read. All the previous comments were then aggregated on one screen, where questions

were asked pertaining to their perception of the factor the treatment reflected, their

privacy concerns for the organization or website they just read reviews for, as well as the

you agree to disclose your personal information to LTPC?”. Subjects answered each

(2010), since the subjects would be processing the information about the organization or

website they have just been exposed to (Dimoka, 2012). After answering all questions,

fMRI experiment of trust and distrust when buying online. Subjects iterated this process

for conditions 1 to 4.

For condition 5, measuring the neural correlates of personal interest, subjects were

product category they were highly interested in obtaining online or an e-service (such as

similar set of questions as in the first four treatments. Since this study is not

contextspecific (i.e. ecommerce, ehealth), and is more concerned about individuals’

decisions to disclose their personal information, the review profiles for simulated

comments from Google Play, where users can rate an application on a five-point scale

application. This is also similar to Amazon reviews. The mean score of users’ ratings are

4.4.2. Experiment 2: 2x2x2x2 Factorial Experiment

and the interaction effects of the independent variables against the dependent variables.

10 High Low High Low

11 High Low Low Low

12 Low Low Low Low

one experiment. This made it easier for them to complete each experiment without having

to reschedule a meeting to do the other experiments. As such, the first eight conditions of

the 2x2x2x2 factorial experiment followed on directly where experiment 1 ended. In so

doing, the choice of product category obtained online or e-service subjects had the most

interest in (condition 5 of experiment 1) was used as the means of manipulating the first

eight conditions of experiment 2 for having a high level of personal interest. Subjects

were asked to keep their choice in condition 5 of experiment 2 in mind, while proceeding

to the next eight conditions, as if the website or organization offered the interested

product category or service of choice.

Subjects were then exposed to the first eight conditions, each of which consisted

concerns and willingness to disclose their personal information to the organization or

that reflected high risk, low distrust and low trust (conditions 3 and 11) had a slightly

higher mean score than the profile with high risk and high distrust, but low trust

(conditions 2 and 10), since the former profile consisted of more negative comments.

Between each condition, subjects were asked to select a number between 1 and 9

before proceeding to the subsequent conditions. This acted as the control condition to

reduce any carryover effects. When subjects completed the first eight conditions, they

were presented with a similar screen as that of condition 5 from experiment 1, but instead

of choosing a product category or e-service they were most interested in, they chose the

one that they were least interested in. This acted as the manipulation for a low level of

personal interest. Subjects then proceeded to complete the remaining conditions 9-16 in

the same manner as they did for conditions 1-8. Conditions 9-16 utilized the same

mentioned above, uncertainty may consist of doubt for the outcome of a situation, but is

differentiated from ambiguity in that it is more negative. As such, it was not included as a

factor in experiment 2, since a review profile that contains comments over whether an

organization or website is risky, or should be trusted or distrusted, should help better

inform individual’s decisions and would not be representative of uncertainty. However,

uncertainty must be tested with personal interest to observe if there are any interaction

effects. Specifically, while the research model in Chapter 3 did not hypothesize a

significant relationship between personal interest and uncertainty, it is incumbent to test

for any effects that may exist between the constructs when conducting an experiment.

Therefore, a 2x1 factorial design (two levels for personal interest, one level for

uncertainty) would be used.

made for condition 5 of experiment 1 to elicit a high level of personal interest. Subjects

were asked to keep this choice in mind while reviewing the profile for the first condition

of experiment 3. Similar to experiment 1 and 2, each comment was read by the subjects,

experiments 1 and 2, and were all rated on a seven point Likert scale.

Subjects were asked to select a random number between 1 to 9 before moving on

4.4.4. Summary of Experiments

The study utilized three within-subjects experiments. However, all three

experiments were conducted as if it were one experiment for the convenience of the

subjects. Experiment 1 consisted of five conditions, used for assessing the neural

correlates of privacy risk, trust, distrust, uncertainty and personal interest, respectively.

therefore sacrificing the external validity of the experiment. However, this tradeoff of

external validity for strong internal validity is necessary in experimental research when

the investigators wish to better understand a phenomenon, by examining the factors that

influence the phenomenon (Sekaran & Bougie, 2013), such as in the case of this study,

where a better understanding of the privacy paradox is the end-goal.

epoched data files were transformed into .slor files which produced sLORETA images,

current source density and localization for each timeframe per file, per subject.

LORETA calculates the volume elements (voxels) of the cortical gray matter

threeconcentric shell spherical head model co-registered on a standardized MRI atlas with

atlas of the human brain, that maps locations of brain structures from individual

differences in the size and shapes of the brain (Talairach, 1998). It consists of three axes,

the x-axis which corresponds to the left and right sides of the brain, y-axis which

used for this study, utilizes the MRI atlas from the Montreal Neurological Institute

Pascual-Marqui (2002) argued that localization error cannot be improved beyond

sLORETA.

were exposed to neutral facial affect. Hotspots (i.e. areas of brain activity) were captured

and compared between each technique, where all brain areas had an equal probability of

containing a hotspot (Pascual-Marqui et al., 2002). Specifically, there was no favored

brain area, and the experiment was used to determine where statistically significant brain

activity occurred in the seventeen subjects. Similarly, single group zero-mean t-tests

utilizing Statistical non-Parametric Mapping (SnPM) were used for each of the conditions

of experiment 1. Essentially, the group means of brain activity were compared against a

group with brain activity of zero-mean.

EEG data transformed to sLORETA images produce localized brain activity for

axes) are multiplied by the total number of voxels (i.e. 3x6329), resulting in 18,987

cannot determine statistical significance for testing brain activity, as multiple

comparisons of nearly one million variables (in this case, based on the data produced

Thus, SnPM solves statistical errors used for testing neural activity by estimating the

probability distribution through the randomization procedure, while retaining the highest

possible statistical power (Pascual-Marqui et al., 2002).

For each condition, all subjects per condition were tested using SnPM, where

onetailed tests were produced, as well as t-statistics for each timeframe. This was done by

t-statistics for each timeframe were determined through randomization tests of 5000

(Nichols & Holmes, 2001). Timeframes with a t-statistic meeting or surpassing the

threshold values at p ≤ 0.05 were group-transformed into sLORETA images for the brain

transformation of data is one technique of many data transformation techniques that are

used to simplify complex data, as well as provide normality to otherwise non-normal data

(Osbourne, 2002). As violations of normality could hinder certain parametric tests, data

transformations are often used in fields such as social sciences (education, psychology)

and biology, where normality is rare (Micceri, 1999; Osbourne, 2010).

ROI analysis has been used in fMRI research (Poldrack, 2006), and was used by

predicting price premiums. Additionally, Dimoka (2010) then developed a regression

model with independent variables based on ROI values to predict price premiums.

Following such method of analysis, the average log CSD per ROI was extracted for each

each condition, and the response of each subject in each condition for privacy concerns

were regressed against their willingness to disclose their personal information.

The regression models produced unstandardized beta coefficients of each

variable based on one unit change in the independent variable, y (Sekaran & Bougie,

while a negative beta coefficient represents a decrease in y when xj increases by one unit.

Similarly, for each beta coefficient, standardized beta coefficients were produced,

whereby the beta coefficients are standardized to a mean of zero and a standard deviation

hypotheses 6 to 11 (both experiment 2 and 3) were rejected or accepted for main effects.

All analysis was done using SPSS. Interaction effects (hypothesis 12) were tested using

significance).

The regression models produced a coefficient of determination called the

Rsquared value denoted by,

to which data points vary when compared to their mean). The coefficient of determination

explains the extent to which the variation in the independent variables vary from the

dependent variables (Sekaran & Bougie, 2013). The adjusted R-square value was also

calculated, denoted by

variables in the model.

An Analysis of Variance (ANOVA) test was produced for each regression model to

explain how well the independent variables explain the dependent variables with an

where an F-value significant at p ≤ 0.05 indicates the model’s independent variables are a

better predictor of the dependent variables as opposed to the constant-only (i.e. β0) model.

Finally, the tolerance (equation 5) and variance inflation factors (VIF, denoted in equation

thumb” are given concerning the tolerance and VIF values for detecting multicollinearity.

The higher the tolerance value (i.e. the closer to 1) the less likely multicollinearity exists,

while a lower VIF indicates lower multicollinearity. While some researchers use a

tolerance value as high as 0.70 to indicate multicollinearity (Dimoka, 2010), values as

low as 0.10 are accepted, while VIF values as high as 10 and lower are considered

acceptable (O’Brien, 2007).

Experiment 3 followed the same type of analysis as experiment 2. An ROI seed

file was created using the sLORETA utility for brain areas associated personal interest

and uncertainty identified in experiment 1 (conditions 4 and 5). The ROI seed file was

regression model was developed using the average log CSD per ROI and privacy

4.5.5. Summary of Data Analysis for all Experiments and Hypotheses Testing.

electrode for each subject per condition using paired groups zero-mean SnPM t-tests.

Essentially, the timeframes per subject per condition were compared, where threshold

tdistributions were produced to determine which timeframes were statistically significant.

The .slor transformed files were then analyzed as a group, where the statistically

significant timeframe(s) indicated the localized source of brain activity.

Hypotheses 6 to 9 and hypothesis 11 corresponded to experiment 2. These

areas identified in conditions 1-3, and 5 of experiment 1. The ROI seed file was then used

to extract the log transformation of CSD of each ROI per subject per condition, and its

nearest neighbor voxel. The average log CSD per subject per condition per ROI was used

the relationship of the log CSD of each ROI and privacy concerns as independent

4.6. Summary

Three within-subjects lab experiments using an EEG were conducted to collect

data and test the hypotheses of this study. However, subjects performed all three

experiments as if they were one. Experiment 1 consisted of five conditions, used for

capturing the neural correlates of privacy risk, trust, distrust, uncertainty and personal

interest. Experiment 2 was a 2x2x2x2 factorial experiment, whereby personal interest,

privacy risk, trust and distrust were all varied on two levels. Experiment 3 was a 2x1

factorial experiment where personal interest varied on two levels, but uncertainty

remained at only one level. Review profiles were used for simulated organizations and/or

personal interest was manipulated by asking subjects to choose a product category or

service they were interest in, using ICTs.

dependent variable for the first model, and an independent variable in the second model,

for each experiment.

Results

5.1. Introduction

This chapter presents the results of the experiments used to test the research

model and hypotheses. Section 5.2 describes the outcome of the pilot study, while section

5.3 contains the results of experiment 1. Sections 5.4 and 5.5 present the results for

experiment 2 and 3 respectively. Section 5.6 contains the summary, included in which,

highlights the hypotheses that were rejected and supported.

Five subjects were used for the pilot study, whereby three were males and two

were females. The primary objective of the pilot study was not used for acquiring neural

correlates, as the sample size was too low for any significant results, nor for examining

relationships between variables used for the regression models in experiment 2 and 3.

The pilot study was first used to identify limitations in the experiments themselves (i.e.

initially used the actual CSD values. However, this led to a lot of outliers, skewed

datasets, and non-normal distributions. Moreover, when actual CSD values were not

computed well in the regression models, which led to almost all variables being removed.

0.05, while a t-value of 4.393 is significant at p-value 0.05 for two tailed tests. There was

significant activation in left BA 40, inferior parietal lobule (MNI coordinates: -40, -50,

60; t-value: -4.20; p < 0.05), and left BA 37, middle temporal gyrus in the temporal lobe

(MNI coordinates: -60, -65, 5; t-value: -4.34, p < 0.05). Thus, hypothesis 5a was not

supported; however, hypothesis 5b was supported and localization for personal interest

was inferred to be correlated with neuronal activity at left BA’s 40 and 37. Figure 13

and coronal planes.

neural correlates of personal interest, risk, trust and distrust was created. This seed file

was used to extract the log-transformed CSD values for each ROI voxel and its nearest

neighbor, for each subject for each of the sixteen conditions. The average log CSD per

ROI was then plugged into a regression model to determine the effect of each neural

correlate in influencing privacy concerns and in a second model to determine the

influence of each neural correlate in influencing personal information disclosure. Table 4

shows the descriptive statistics for the variables of the sixteen conditions.

Right BA32 5.21 1.19

Left BA 9 5.68 1.17

Right BA 47 5.98 1.10

Right BA 31 4.45 1.21

Right BA 13 5.49 1.17

Left BA 40 4.99 1.17

Left BA 37 5.32 1.04

Personal Information Disclosure 3.03 1.92

5.4.1. Regression Model for Privacy Concerns

adjusted R-square value was 0.036. This indicated that the variance predicted by the

independent variables was poor. However, the model had an F-value of 2.852 at p =

0.007, indicating its significance in predicting the dependent variable, privacy concerns,

variable privacy concerns.

Independ- Beta Standardiz- Significan- Signific- Toleran- VIF ent

Coeffici- ed Beta ce (t-value) ance (p- ce Variables ent Coefficient value)

The results of Table 5 indicated that only right BA 13 is significantly related to

privacy concerns. All other variables are above the threshold 0.05 p-value. However, BA

13 has a standardized beta coefficient of -0.294, indicating a negative relationship with

privacy concerns. Essentially, higher neural activity in BA 13 is related to lower privacy

concerns. This result is opposite to hypothesis 8a. Thus, while the relationship is

significant, the hypothesized direction of the relationship is not supported. Similarly,

hypothesis 6a was not supported since the relationships are not significant. There were

also no significant relationships when checking the interaction effects of the independent

5.4.2. Regression model to Determine the Relationship between Risk and Trust

The neural correlates of risk (right BA 32 and left BA 9) were regressed against

the dependent variable of trust (right BA 47). The regression model contained an

Rsquared value of 0.799 and an adjusted R-squared value of 0.798, indicating a high

power in explaining the variance of the dependent variable with the variance of the

independent variables. The model produced an F-value of 693.95 at p < 0.001. This

indicated the independent variables were better predictors of the dependent variable than

below 10, indicating that multicollinearity may not have been a problem in this model.

Independ- Beta Standardiz- Significan- Signific- Toleran- VIF ent Variable Coeffici- ed Beta

ce (t-value) ance (p- ce

Right BA 32 0.960 -1.039 -18.042 0.000*** 0.174 5.76

Left BA 9 -0.153 -0.163 -2.822 0.005** 0.174 5.76

5.4.2. Regression Model for Personal Information Disclosure

The dependent variable, personal information disclosure was regressed against the

independent variables, Right BA 32, Left BA 9, Right BA 47, Right BA 31, Right BA 13,

Left BA 40, Left BA 37, and privacy concerns. The regression model had an R-square

the independent variables significantly predicted the dependent variables. Table 7

displays the regression coefficients of the independent variables on personal information

disclosure.

Right BA 31 0.084 0.053 0.489 0.646 0.113 8.85

Right BA 13 -0.413 -0.253 -2.532 0.012* 0.151 6.61

Left BA 40 -0.169 -0.103 -1.120 0.227 0.209 4.78

Left BA 37 0.201 0.109 2.097 0.037* 0.561 1.78

Concerns

* p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001

Apart from the right BA 31 and left BA 40, all Brodmann Areas were significantly

related to personal information disclosure. As indicated in the results from experiment 1,

privacy concerns are related to brain activity in both the right BA32 and left BA 9. Right

BA 32 has a significant and negative relationship with personal information disclosure

(standardized β = -0.228, p < 0.05), indicating that an increase in brain activity in the

right BA 32 is related to a decrease in an individuals’ decision to disclose his/her personal

information. The results of BA 32, when regressed on personal information disclosure

partially supports hypothesis 6b. However, left BA 9 has a positive relationship with

0.05), whereby an increase in an individual’s decision to disclose personal information is

related to an increase in brain activity in the left BA 9. Similarly, right BA 47, the

level of an individual’s personal interest, and both had positive beta coefficients,

1.68 2.045 0.0434*

-9.28 -2.241 0.0271*

disclose personal information. However, only BA 37 was significant, with a standardized

β = 0.109. Hypothesis 9 was thus partially supported. Privacy concerns, were negatively

and significantly related to personal information disclosure, with a standardized β = 0.683

(p < 0.001), indicating an increase in privacy concerns will cause a decrease in personal

information disclosure, supporting hypothesis 11. Tolerance and VIF values were

collected for each variable, whereby each variable, apart from BA 32 were above the 0.10

threshold for tolerance, and below the 10 threshold for VIF indicating multicollinearity

may not be present.

Table 8 displays the interaction effects of independent variables on the dependent

independent variables.

Interacting Variables Beta Coefficients Significance Significance

0.072 and an adjusted R-square of 0.024. These values were low in predicting the

0.438 and an adjusted R-squared of 0.364 was achieved suggesting a fair degree of

neural correlates of uncertainty and personal interest on either privacy concerns or

personal information disclosure. Table 9 summarizes the results for the privacy concerns

regression model, while Table 10 summarizes the results for the personal information

disclosure regression model.

Variables Beta

Standardized

Significance

Significanc-e

Tolerance VIF

perceptions are associated with right BA’s 40 and 47, and left BA’s 40 and 37,

respectively. Experiment 1 supported hypotheses 1c, 3b, 4a and 4b.

For experiment 2, the average log CSD of ROI for each brain area identified in

model with the dependent variable of privacy concerns. The results indicated the model

However, this result is opposite to what was predicted in hypothesis 8a. Hypothesis 7a

was assessed by developing a regression model for the dependent variable of the neural

correlate trust (left BA 47) against the independent variables of the neural correlates for

partially supported, while hypotheses 7b and 11 were fully supported. Also, there was a

against privacy concerns and personal information disclosure, there were no significant

main or interaction effects, thus hypothesis 10a and 10b were rejected.

6.2.5. Privacy Concerns

However, data collected for privacy concerns utilized a one-item survey question of

selfreported data. Essentially, privacy concerns, like privacy risk, have been found to be

insightful findings. Moreover, neural correlates of privacy concerns were not assessed,

which may have provided useful insights as to how the human brain processes

information privacy.

statistically significant, and had the strongest effect when compared to all the other

relationship is consistent with findings in the information privacy field, where privacy

concerns negatively impact the use of ICTs that require individuals to disclose their

6.2.6. Interaction Effects

dependent variable of personal information disclosure (see Table 8 in Chapter 5). In total,

only three out of twelve interactions that had a positive relationship with personal

Alternatively, all the significant interactions with negative relationships had a beta

coefficient of -1.33 as its lowest, but -9.28 as the highest, which was an interaction

trust were present predicted a reduction of personal information disclosure (i.e. negative

beta coefficients). These findings were similar to Dimoka (2010), which found distrust

was more salient than trust in predicting price premiums. Additionally, the high number

of interactions of the independent variables and their predicted negative relationships and

magnitude (i.e. beta coefficients) with personal information disclosure reinforces the

weight negative outcomes have over positive ones, when individuals make decisions

regarding rewards and losses (Kahneman & Tversky, 1979).

6.3. Contributions

There were four contributions from this study to information privacy research.

Firstly, this study provides a better explanation to the privacy paradox in that individuals’

intertwine with one another. The findings thus supports the research argument that

contribution, where the findings of cognitive neuroscience was applied to the extended

privacy calculus model to address its limitation in assuming individuals are rational

using sLORETA technique to identify the neural correlates of mental processes posited to

be involved in explaining the privacy paradox, and using SnPM for the analysis. These

6.3.1. Research Contributions

emotional mental processes. The brain areas identified as neural correlates to the factors

such as privacy risk, trust, distrust, and personal interest are responsible for

privacy concerns, and personal information disclosure. This indicates that individuals’

privacy-related decisions are neither purely emotional nor rational, as there exists

interconnectivity between these brain areas as suggested by Phelps (2006). Essentially,

the privacy paradox cannot be explained solely through a logical cost-benefit analysis or

The constructs of privacy risk, trust, privacy concerns and personal interest in the

extended privacy calculus model were modeled as enduring (institutional) beliefs

individuals have in the context of information privacy, which influences individuals’

decisions to withhold or disclose their personal information. Yet, these constructs have

been found to be correlated with neural activity in specific regions of the brain when

individuals are processing stimuli, i.e. they are produced based on situations and are more

a ‘state-of-mind at a given time’ as opposed to an enduring belief (Dimoka et al., 2007;

2011; Sur & Sinha, 2009). Additionally, studies have found that the activation of brain

areas when individuals are in a specific situation influences behavioral outcomes

(Dimoka, 2010; Vance et al., 2014). As Dimoka et al. (2011) indicated, the use of

neuroscience can advance the IS field. One such opportunity occurs where antecedents of

The findings of this study fulfill this opportunity highlighted by Dimoka et al., (2011)

where specific neural correlates, that included emotional responses and rational

processes, significantly influenced individuals to disclose or withhold their personal

developed by Dinev and Hart (2006). The extended privacy calculus model is limited in

extended privacy calculus model were mapped to brain areas of neural activity using the

correlates of uncertainty did not have any significant relationships with privacy concerns

or personal information disclosure, there were significant relationships with the neural

correlates of privacy risk, trust, distrust and personal interest to personal information

disclosure. Essentially, this study establishes distrust as an integral factor in the extended

privacy calculus model, and explaining the privacy paradox. This study also advances the

IS field as IS constructs were mapped to specific brain areas, which would allow for

better understanding the nature and dimensionality of these constructs as opposed to

utilizing the metrics of self-reported data that would be inadequate in assessing and

The third contribution of this study was related to the research methodology, and

contributed to both the information privacy field, as well as NeuroIS in general. This

study used (standardized) low tomography brain electromagnetic tomography to identify

the neural correlates of mental processes posited to be involved in explaining the privacy

(Pizzagalli, Oakes, & Davidson, 2003), and hippocampal regions.

While there is a growing body of literature in the neuroscience field that utilizes

time of writing this dissertation, the use of sLORETA seems very rare in IS, with the only

LORETA”, and “eLORETA”) did not return any results on databases such as ProQuest,

ABI/InFORMs, and Web of Science for sLORETA-based research in IS, much less

information privacy. The only result obtained from searching these databases with regards

to neuroIS and information privacy returned an editorial calling for the use of

neuroscience techniques in information privacy to better understand privacy constructs

Essentially, this study contributed by answering such a call to use neuroscience

techniques to better understand the privacy paradox, while also contributing to NeuroIS

by using sLORETA, a technique for identifying brain areas with high temporal resolution.

Other than the analysis of the localization for key brain areas, sLORETA provides

additional advantages such as the ability to provide analysis of frequency domains

involved in specific tasks, along with localization (i.e. origination) of these frequency

domains (Cannon, 2012; Massar et al., 2012). Furthermore, functional connectivity

dependency of neuronal activation patterns of anatomically separated brain regions”

(Lang, Tome, Keck, Gorriz-Saez, & Puntonet, 2012, p. 1).

An additional advantage of using sLORETA for identification of key brain areas

collecting data drastically. Traditional neuroimaging tools such as fMRI and PET are very

respectively (Riedl et al., 2009). EEG devices, in comparison, are much cheaper than

purchasing fMRI and PET machines, while using the equipment from third parties

particularly in neuroIS and information privacy, since fMRI and PET studies have smaller

use of EEG devices could provide more forms of analysis (i.e. frequency analysis, event-

decisions. Research in NeuroIS is limited, with a number of researchers appealing to the

use of cognitive neuroscience to better understand information privacy (Dimoka et al.,

2007; 2011; Riedl et al., 2009). Statistical Parametric Mapping (SPM) using general

linear models is an often used approach of analysis for functional neuroimaging data,

even in the fields of neuroscience and neuropsychology (Dimoka, 2012; Nichols &

Holmes, 2001). In NeuroIS fMRI research, Dimoka (2010) used SPM to test for

significant brain areas associated with trust and distrust. However, Nichols and Holmes

(2001) explained that SnPM can surpass SPM for analyses of brain data with low degrees

(SPECT).

organization’s survivability and growth depends greatly on insights and predictions they

gain from data mining techniques. This data allows for them to understand patterns and

perceptions of an organization by society. As was evident in the case of ChoicePoint, this

Organizations should therefore handle the personal information they collect from

perceptions have more weight than positive ones. Thus organizations should reduce

2015). Additionally, by acting in a manner that suggests an organization’s ethical

disposition as morally altruistic, with high concern and practices reflecting a “proactive

privacy” attitude could lead to increased trust between clients and organizations.

Organizations could achieve this by creating and often reviewing comprehensive privacy

policies, developing a culture of privacy within an organization, designing informative

and easy to read privacy statements, and helping individuals that were affected by privacy

Tejay, 2015). Organizations should also invest in messages, slogans, and cues that

highlights the importance of clients’ information privacy. As can be seen in cognitive

trusting perceptions and developing an environment of trust should impact individuals’