The Role of Trust in Feelings in Confirmation Bias

Master Thesis


MSc Marketing Management

The Role of Trust in Feelings in Confirmation Bias

By:

Stephanie Aderika Sihombing

Date of Submission: 17

^th

of June 2019

The Role of Trust in Feelings in Confirmation Bias

University of Groningen Faculty of Economics and Business

MSc Marketing Management Master Thesis

First Supervisor: A. Schumacher, MSc.

Second Supervisor: prof. dr. B. M. Fennis Completion Date: 17/06/2019

Stephanie Aderika Sihombing S2936712

Hofstraat 10A

9712JB, Groningen, The Netherlands 
+31619514767

s.a.s.sihombing@student.rug.nl

Abstract

This research examined how different levels of trust in feelings affect individuals’ tendency to distort information upon the decision-making process. Moreover, this research also examined the moderating effect of counterarguing mindset on individuals’ tendency in distorting information under different levels of trust in feelings. Within this research, we propose that individuals with high trust in feelings are more prone to distorting information.

A 2 (trust in feelings: high versus low) x 2 (counterarguing mindset versus control condition) between-subject design online study through Amazon Mechanical Turk was conducted to test our hypotheses. According to the results, there are no significant effects found in our research. Both the main and second hypothesis are not supported, and there is not enough proof to reject the null hypothesis. Therefore, information distortion levels in the decision- making process are not affected by the level of trust in feelings within individuals.

Furthermore, with differing levels of trust in feelings, the level of information distortion is not increased or decreased with counterarguing mindset.

Keywords: trust in feelings, confirmation bias, information distortion, consistency mindset,

counterarguing mindset

Acknowledgement

With this, I present my master thesis to close the chapter of my Master program in Marketing Management. Even though this thesis only took five months to be completed, I am proud to say that it is the hard work of my four and a half years of bachelors and one-year masters education. I want to use this platform to thank my thesis supervisor, Anika Schumacher, MSc., for her help and guidance throughout my master thesis journey. She gave detailed, constructive feedbacks, which helped me in improving my thesis with ease. Furthermore, I would like to thank my second supervisor, prof. dr. Bob M. Fennis, who is willing to evaluate my thesis and taking the time to do so.

I would also like to thank my friends as we have helped each other in improving our thesis

contents, proofread my thesis, and encouraged me to do my best for my thesis. I would like to

thank my boyfriend, who has helped me through analysis, motivated me, and for believing in

my capabilities. Last but not least, I would like to thank my family back in Indonesia for their

endless support and love for me.

Table of Contents

ABSTRACT ... 3

ACKNOWLEDGEMENT ... 4

INTRODUCTION ... 6

LITERATURE REVIEW... 8

T

RUST IN

F

EELINGS

... 8

C

ONFIRMATION

B

IAS

...11

T

RUST IN

F

EELINGS

& C

ONSISTENCY

M

INDSET

...12

C

ONSISTENCY

M

INDSET

& C

ONFIRMATION

B

IAS

...13

METHODOLOGY ... 14

S

AMPLE

...15

S

TUDY

D

ESIGN

...16

M

ANIPULATING

T

RUST IN

F

EELINGS

...16

C

OUNTERARGUING OF

C

ONSISTENCY

M

INDSET

...17

P

RODUCT

E

VALUATIONS

...17

RESULTS ... 19

M

ANIPULATION CHECK

...19

M

AIN

A

NALYSIS

...20

DISCUSSION ... 22

LIMITATIONS ... 23

IMPLICATION ... 25

FUTURE RESEARCH ... 26

CONCLUSION ... 26

REFERENCES ... 28

APPENDIX... 32

A

PPENDIX

A ...32

A

PPENDIX

B ...49

Introduction

Information availability and diversity in recent years can be easily accessed by the public through the internet and other news and social media platforms. Nevertheless, amidst these varieties of information, individuals still engage in confirmatory bias; otherwise known as information distortion. It is a biased evaluation of given alternatives and favouring the leading option to fit one’s own prior beliefs (Russo, Meloy, and Medvec 1998). On a regular basis, consumers exhibit preference bias upon choosing products or services (Chaxel and Han 2018; Russo, Meloy, and Medvec 1998). Consumers create this form of information distortion in order to achieve consistency so that the incoming information would align with their emerging preferences (Chaxel, Russo, and Wiggins 2016; Russo et al. 2008). To simplify, consumers are unconsciously interpreting new information in methods that will favour their own superior option or believing that information given to them favours one of the two options (Chaxel and Han 2018).

It should be noted that information distortion occurs subconsciously, and consumers find it hard to counteract its effects (Russo and Yong 2011). Drawbacks of information distortion range from inaccuracy in judgements, to the risk of ending up with an inferior alternative (Russo, Meloy, and Medvec 1998). By distorting information, consumers would be more confident that the leading option is better than the others. Therefore, leading to consumers refraining from seeking out further information about other alternatives.

Aside from consumer behaviour, other researches have proven that information distortion exists across various situations. These studies include examinations of medical decision making (Kostopoulou et al. 2012; White, Wearing, and Hill 1994), information distortion by auditors and salesperson (Russo, Meloy, and Wilks 2000), biased interpretation of evidence by prospective jurors (Carlson and Russo 2001), biased boxing verdicts by professional boxing judges (Tyszka and Wielochowski 1991), biased expertise in horse track bettors (Brownstein, Read, and Simon 2004), and examining determinants of information distortion (Beckmann and Kuhl 1984).

Based on the information above, we propose that individuals’ general level of trust in feelings

is one of the factors that might activate confirmatory bias. People are inclined to trust their

feelings more if they have a history of experiencing numerous successes in executing

decision-making situations rather than those who frequently garnered negative experiences in decision making with their gut feeling (Avnet, Pham, and Stephen 2012). According to Avnet et al. (2012), “trust in feelings” refers to a situation whether or not an individual believes that in general, their feelings could point out towards the correct direction in decision making processes or judgements. Research regarding trust in feelings pointed out that the use of feelings is more prevalent and relevant upon judging targets that seem to be appropriate to assess; in the sense of conveying characteristics of particular objects (Pham 1998). Such objects are the ones that elicited hedonic or experiential motives, and it induces consumers to engage purchase or preference decision with their feelings rather than with their cognitive abilities (Pham 1998).

Research has also shown that trust in feelings is changeable rather than a set of an individual’s fixed trait. By changing the perceived success history of a person upon reliance on their own feelings as a decision-maker, subsequently, there will be a shift in one’s perception of trust in their own feelings. The procedure called ease-of-retrieval effect manipulates individual’s perceived success history upon reliance on feelings in decision- making processes (Schwarz, Bless, and Bohner 1991)

Prior research has shown that trust in feelings activates consistency mindset (Lee, Amir, and

Ariely 2009), and according to Chaxel et al. (2018), a consistency mindset results in

confirmation bias. Yet, no research investigated the influence of trust in feelings on an

individual’s likelihood to engage in confirmatory reasoning. The purpose of this research is to

bridge this gap. Specifically, we propose that this is an important gap as this might open a

new way for marketers to avoid or reduce choice bias within consumers’ minds when faced

with a decision-making process. Moreover, we would like to observe the effect of priming

counterarguing in the minds of participants. Priming a counterarguing mindset – the act of

refuting several propositions prior to the the main experiment – will disturb a consistency

mindset and thus, reduces information distortion (Chaxel and Han 2018). This procedure

applies to various domains such as decision making in a subsequent setting or judging certain

objects (Xu and Wyer 2012). We expect that by activating counterarguing mindset,

individuals would rely less on their feelings, and in turn, decreases confirmation bias.

Therefore, the proposed research question is as follows: Are individuals with high trust in feelings (compared to those with low trust in feelings) more likely to engage in confirmatory bias?

Two hypotheses in our paper is elaborated in detail on the upcoming section. Our experiment was conducted in an online study setting through the usage of Amazon Mechanical Turk (MTurk). According to our result, we have failed in proving our two hypotheses; thus, yielding an ambiguous answer to our research question. Based on our result, there are no significant differences between high trust in feelings versus low trust in feelings conditions and their likeliness of engaging in confirmatory bias. There was also no significant difference when these two conditions were moderated with a counterarguing mindset towards the likeliness of engaging in confirmatory bias.

This paper is constructed as follows: First, the literature review section will elaborate in detail regarding trust in feelings, confirmation bias, and the interactions of our model. Furthermore, two hypotheses are constructed based on these theories, and a conceptual model of our hypotheses are formed. Afterwards, the methodology section will explain our experimental setup, followed by the result section, where our statistical results are explained. Then, there will be a discussion section to analyse our overall experimental results. Finally, some limitations, practical implications, and ideas for future research will be provided.

Literature Review

Trust in Feelings

The role of trust in feeling is revolutionary as it gives a whole new meaning towards the importance of affect, emotions, judgement, and feelings within decision-making processes (Pham, 2004). We propose that some people perceive their feelings as more or less

“trustworthy”, while others might perceive their feeling as “untrustworthy”. Perceiving one’s feelings as trustworthy means that it tends to point to the right direction or choice.

Meanwhile, perceiving one’s feelings as untrustworthy means that it tends to point to the wrong direction instead. There are two reasons why individuals behave in such a way:

According to Avnet et al. (2012), an individual’s past success or failure of relying on their

feelings in decision-making processes will contribute to the evaluation of trustworthiness in

one’s feelings. Moreover, the social and cultural environment of individuals will also contribute to the trust of their own feelings. Some cultural environments encourage the usage of feelings in decision-making through norms and values, while others discourage such practices (Avnet, Pham, and Stephen 2012). Therefore, trust in feelings can be seen as a flexible recurring habit as it might change depending on whether or not different cultural norms and values are internalized, or certain changes regarding the perceived history of success or failure occur (Avnet, Pham, and Stephen 2012).

When individuals are faced with a decision-making or judgement task, it is also important to note that every individual differs in their belief regarding the reliability of their own feelings regardless of the relevance and observed representativeness of the task (e.g. choosing a product when shopping) (Avnet, Pham, and Stephen 2012). Some individuals are more likely to believe that their gut feelings guide them in various decision-making processes (Salovey et al., 1995). The theory of affect-as-information showed that individuals are selective in their reliance on feelings in a decision-making process because it depends on the observed informative value of the feelings (Schwarz and Clore 1983). Nevertheless, it does not imply that people consciously think about how representative their feelings are toward a target (e.g.

goods/services). By default, people made assumptions on whether or not their feelings represent the target in question (Schwarz, Bless, and Bohner 1991).

Feelings are much more salient when individuals are instructed to focus on them (Siemer and Reisenzein 1998; White and McFarland 2009). A process called “affect-referral” (Wright 1975) happens when individuals rely on previously stored assessments in memory. This process helps individuals in forming final judgements of current goods or services at hand (Lingle and Ostrom 1979; Lynch, Jr., Marmorstein, and Weigold 1988). Sometimes, consumers would instead rely on feelings experienced in past encounters to judge a target rather than weighing attributes owned by the target itself. This situation occurred when they perceive feeling as a reliable source of information (Schwarz, Bless, and Bohner 1991).

However, feelings would not be utilized if their informational value is irrelevant to the target (Pham 1998). Moreover, individuals are generally not good at accurately recalling past feelings or emotions because often they would rely on inaccurate intuition when recalling and imagining affective states that they are not currently experiencing (Blumberg et al. 1998;

Loewenstein 1996; Snell, J., Gibbs, B.J. & Varey 1995). According to the “affect-as-

information”, if an individual feels positive about the target, it implies that they like it, and if

they feel negative towards the target, it implies that they should not like it. Positive feelings are subconsciously allowing less cognitive effort in decision-making processes, while negative feelings induce individuals to put more effort, for example, in assessing attributes and reduce people’s reliance on heuristic cues. This is due to the fact that positive feelings generated a sense of safety, while negative feelings generated a sense of threat (Schwarz, Bless, and Bohner 1991).

There are several possible consequences of trust in feelings on consumer behaviour. Firstly, a momentary high level of trust in feelings will induce individuals in choosing targets that may seem affectively appealing, even when other alternatives are statistically better in terms of attributes (Avnet, Pham, and Stephen 2012). This means that consumers might lose the opportunity to obtain the best goods or services just because they chose what “felt right” for them at the moment. According to the “jelly bean” paradigm, consumers who scored high on trust in feelings are more inclined to choose options that are affectively appealing but statistically inferior target rather than the less affectively appealing but statistically superior target (Denes-Raj and Epstein 1994). Secondly, it is easier to influence consumers over advertisements when they exhibit a high trust in feelings because they rely more on their feelings toward the advertisement rather than consumers who exhibit a low trust in feelings (Avnet, Pham, and Stephen 2012). Thirdly, consumers would reject unfair offers more under high trust in feelings. This happens because the dominant emotional feedback would be irritation or resentment towards the unfair offer, rather than a materialistic mindset that forces the acceptance of the offer (Avnet, Pham, and Stephen 2012). Nevertheless, high trust in feelings within individuals does not increase or decrease the tendency of accepting fair offers.

Cognitive consistency refers to the consistency of beliefs wherein there is a matching fit

between newly obtained knowledge and one’s current feelings (Chaxel, Russo, and Wiggins

2016). In another research, cognitive consistency is a way to refute a particular hypothesis or

information (Kruglanski and Shteynberg 2012). Aside from that, affective feelings administer

cues within socially consistent individuals. People would regard their feeling as trustworthy

when their personal consistency emanate reliability and stability in regards to the

environment when used as a ground for judgement (Pham 2004). The fact that affect, as a

decision system, is a residue of individuals’ historical occurrences means that people have the

ability to group targets and objects in a consistent fashion (Pham 2004).

Confirmation Bias

Research has shown that when consumers evaluate incoming information to choose certain goods and services, they tend to exhibit a confirmation bias; otherwise known as information distortion. Consumers often evaluate new information in a way that it would support their existing belief and therefore, would favour the leading brand (Russo, Meloy, and Medvec 1998). They are biased towards their brand preference when they had a positive historical correlation between themselves and the brand (Russo, Medvec, and Meloy 1996). The reason why consumers distort information is due to the need for preserving cognitive consistency between old and new information (Russo, Meloy, and Medvec 1998; Russo et al. 2008). It is essentially a coherent belief that information given to a consumer favours one over the others and the belief that one information is better than the others.

The trade-off for distorting information is the possibility of choosing an inferior option due to the reduction of accuracy (Russo, Meloy, and Medvec 1998). After individuals commit to a particular belief or bias, they will collect supportive information and disregard unsupportive ones in order to avoid cognitive dissonance (Festinger 1962). Cognitive dissonance is the discomfort due to the presence of incongruous or dissonant cognitions. When individuals encounter a challenge to their beliefs or attitudes, they would then try to reduce cognitive dissonance. This, in turn, would enhance information distortion within an individual (Beauvois, Joule, and Joule 1996; Festinger 1962). Pre-decisional distortion may occur even when there are no prior preferences toward a product/target. According to Russo et al. (1998), the magnitude of pre-decisional distortion is more substantial than those of post-decisional distortion because of the reduction of cognitive dissonance.

The tendency of individuals to choose favourable information over unfavourable information

is more prominent when issues are high in value relevance (Festinger 1962). If unfavourable

information is high in utility or if there is no goal in sight, confirmation bias will be smaller

in effect (Hart et al. 2009). On the bright side, information distortion may help consumers in

deciding over ambiguous product information. Consistency of product evaluations with

consumers’ historical preferences is used as a heuristic guide in order to obtain accurate

choice (Russo, Meloy, and Medvec 1998).

The prevalence of information distortion can be attributed to several things. Some individuals are very confident and have strong convictions that their preference is the best option there is (Russo and Yong 2011). This distorted information may turn out to be too optimistic and impair these individuals from risk judgement. Moreover, research has shown that individuals who rely on their feelings more in decision-making processes are less prone to cognitive dissonance in comparison with those who utilize reason-based judgements (Lee, Amir, and Ariely 2009). Thus, leading to greater cognitive consistency and preference stability for these high trust in feelings individuals.

Trust in Feelings & Consistency Mindset

Greater reliance on emotional processes will also contribute to a higher level of preference consistency (Lee, Amir, and Ariely 2009). An evaluation based on emotion is considered as much more reliable and consistent due to its holistic nature in comparison to its cognitive analytical counterpart (Epstein 2003). Upon evaluating a subject, emotional processing focuses on the essence of the subject and thus would be more consistent from time to time (Epstein 2003). On the other side, cognitive or analytical processing is more focused on details of the subject, making it sensitive to any changes in the subject at hand; and thus, vulnerable to inconsistencies. Lee et al. (2009) believe that the emotional process provides consistency due to its higher accuracy and speed in evaluating a target.

Consumers, in general, should generally have their own preference structure, consistently assessing various targets in a congruous manner (Lee, Amir, and Ariely 2009). Greater utilization of feelings will generate consistency within individual’s evaluation, whereas a greater reliance on cognitive processing would be more likely to generate inconsistencies.

Therefore, following the theory of (Lee, Amir, and Ariely 2009), individuals who trust their

feelings more should display a consistency mindset in their decision-making processes,

generating consistent preferences. A counterarguing mindset refers to the task wherein

participants will contradict a series of statements provided, and subsequently will reduce the

amount of information distortion in the following task by contradicting their own

preconceptions (Chaxel and Han 2018). Counterarguing mindset will, in turn, disrupt

consistencies in an individual’s evaluations. This method would be further explained in the

methodology section.

Consistency Mindset & Confirmation Bias

When an individual distorts given information in order to favour their preferred outcome, it is called the confirmation bias (Jonas et al. 2001). Individuals tend to seek for confirming information during the decision-making process for their favoured alternative (Hoch and Ha, 1986), and this behaviour is due to the desire to maintain consistency in assessing targets (Russo, Meloy, and Medvec 1998). Research has shown that earlier civilizations seek consistency in their daily life in order to survive and predict their ever-changing environment.

Nowadays, individuals strive for consistency of information and their retained knowledge in order for them to incorporate this new information into their belief mechanism as they wish to maintain a consistent perception of their environment (Gawronski and Strack 2012).

Individuals will attempt to differentiate between available options in the pre-decisional situation in order to evaluate the disparity between their preferred preference and the others (Svenson 1996). When an individual is biased towards a specific target, unconsciously they will restructure the importance of target attributes in such a way that it will favour their favourite choice, and when given additional information regarding targets they will acquire a sense of memory retrieval bias (Svenson 1996). Individuals defend their beliefs and attitudes through selective exposure; by looking for information that is in favour of them and steering clear from those that are prone to challenge their beliefs (Hart et al. 2009).

In overcoming information distortion, we propose the usage of Xu & Wyer’s (2012)

counterarguing mindset framework wherein participants refute several prepositions at the

beginning of the experiment. A counterarguing mindset refers to the task wherein participants

will contradict a series of statements provided, and subsequently will reduce the amount of

information distortion in the following task by contradicting their own preconceptions

(Chaxel and Han 2018). To clarify, “mindset” is a cognitive procedure linked towards a

particular objective. It will then, in turn, influence individuals decision-making process in the

following situation or even in another entirely different condition (Xu and Wyer 2012). By

priming a counterarguing mindset, participants would be induced in more or less negative

affect and therefore is much more alert in evaluating upcoming information in the next stage

of the experiment. There, these individuals would rely less on their feelings upon decision-

making processes. It is important to disrupt information bias in consumers by priming

counterarguing in order to refute their current conceptions and preferences. We speculate that

this counterarguing mindset will affect individuals’ trust in their own feelings (information

distortion), and in turn, would prompt them to reconfirm their existing belief and reduces bias in decision-making processes.

Based on the constructed logic above, our proposed hypotheses are as follows:

• H1: High trust in feelings individuals will be prone to confirmation bias, while low trust in feelings individuals will experience a lower level of confirmation bias.

• H2: High trust in feelings individuals primed with counterarguing mindset will experience a lower level of confirmatory bias, while low trust in feelings individuals primed with counterarguing mindset will experience even a lower level of confirmatory bias.

The conceptual framework of our paper is depicted below:

Figure 1: Conceptual framework

Methodology

Within this experimental section, several tests were conducted in order to test all hypotheses.

The experiment was conducted in a quantitative manner, utilizing the online platform of

Amazon Mechanical Turk (MTurk). In the beginning, the sample used for our experiment

will be discussed. Subsequently, the overall study design will be elaborated, and then

followed by the manipulation of trust in feelings, moderation condition, and lastly their

impact on the final measurement for confirmation bias in product evaluations.

Sample

In order to examine the effect of trust in feelings with confirmation bias and its moderating variable -- consistency mindset -- a survey of 290 men (M

age

= 30.16, SD = 8.371) and 231 women (M

age

= 33.75, SD = 11.562) with Amazon’s Mechanical Turk (MTurk). The obtained sample consisted mostly of American, Indian, and other nationalities (Cumulative Percent = 82.5%). Out of 521 respondents, we excluded 242 participants (47.26% of the total sample) for not answering the survey properly. There are several criteria on why we excluded these participants. Firstly, looking at the essay sections, these participants: a) copy-pasted answers from sources; identified by the existence of citations, and identical sentences in comparison to other participants’ answers. b) failed to understand the context of our first and second essay sections; meaning that their answers deviate from the topic at hand. c) answers were too short; e.g. one word per essay. d) poor usage of English in essay writing. e) gibberish answers; meaning that participants wrote illegible words (e.g. “gfjfcmvhz”), and f) usage of inappropriate words & profanity. The last criteria is the most prominent and easy to point out:

failing the attention check. Most of these discarded participants were suspected to be mechanical bots due to the criteria above.

After excluding those unqualified participants, we were provided with 143 men (M

age

= 32.71, SD = 10.154) and 136 women (M

age

= 37, SD = 13.233) with a total of 279 participants. Most of these participants were from The United States and India, indicated by

“Others” in the nationality check (Valid Percent = 79.2%). The rest of the participants were from Indonesia (Valid Percent = 10.4%), Germany (Valid Percent = 9.7%), The Netherlands (Valid Percent = 3.9%), and Belgium (Valid Percent = .7%).

N %

Participants 279

Gender

Female 136 48.7%

Male 143 51.3%

Nationality

Dutch 11 3.9%

German 16 5.7%

Belgian 2 0.7%

Indonesian 29 10.4%

Others 221 79.2%

Table 1: Participants’ Demographic

Study Design

With the usage of Qualtrics as our survey software, it assigned 279 participants randomly to a 2 (trust in feelings: low trust in feelings versus high trust in feelings) x 2 (counterarguing mindset versus control condition) between-subject design. Participants were told that they are going to take three different studies with the theme of “Consumer purchasing behaviour”. In the first study, participants were asked to write short essays in order to manipulate their trust in feelings condition, which is used to construct the independent variable for this study (Avnet, Pham, and Stephen 2012). In the second study, participants were asked to write a short essay about their opinion regarding two statements. It is expected that participants would refute or counterargue these statements, and would prevent them from distorting information in the next study (Xu and Wyer 2012). Priming counterarguing behaviour is expected to reduce consistency mindset and therefore reduce confirmation bias. Thus, this study is used to construct the moderating variable. In the third study, participants were asked to evaluate two products of the same nature, and create a final decision on which product they preferred more. By the end of the study, participants had to rate how much they trusted their feelings in general. The third study was used to construct the dependent variable.

Manipulating Trust in Feelings

For the purpose of this study, we implemented the manipulation of trust in feelings by Avnet

et al. (2012). This manipulation is based on the ease-of-retrieval paradigm by Schwarz, Bless,

and Bohner (1991). Within this study, participants were asked to retrieve situations wherein

they successfully relied on their feelings; in other words, when their hunches came out to be

true. Participants in the high trust in feelings condition were instructed to describe two

situations wherein they relied on their feelings upon decision-making process, and it turned

out to be the right thing to do. Wherein participants in the low trust in feelings condition were

instructed to describe ten situations. All participants from both conditions were given 4

minutes per essay to complete their tasks. It is expected that participants who had to describe

two situations would find it easier and forming a perception of successful reliance on feelings

in the past. On the contrary, participants who have to describe ten situations would find it

harder to do and thus, formed a perception of failure in past reliance on feelings in decision-

making processes. The original study by Avnet et al. (2012) utilized a 7 minutes timeframe

for this manipulation; however, due to the online nature of our survey, we decided to reduce

the timeframe into 4 minutes, and the page would proceed on its own when the timer runs

out. Moreover, we did not have the opportunity to conduct a pre-test due to the time constraints of this report.

Counterarguing of Consistency Mindset

Following the main manipulation – in what seemed to be an unrelated study – participants were either asked to give out their opinions based on two different propositions that induce counterarguing or asked to give general descriptions of two different topics as a control condition (Xu and Wyer 2012). For the moderating variable, participants were asked to give a

“agree/disagree” opinion in an open-ended format regarding: “Human activity has no major impact on the environment” and “Higher education should not be available to all, but only to a small minority of selected students”. A pre-test had been conducted by Chaxel et al. (2017) that more than 90% of participants disagreed with these propositions. These propositions are the ones that are commonly refuted. Nevertheless, people differ in their own opinions, and it is inevitable to encounter several responses that agree with those propositions. However, our goal here is to instil a reduction of consistency mindset through counterarguing. Hence, decreasing the reliance of feelings by each participant as a basis for judgements. It was predicted that participants in the high trust in feelings condition would experience a lower level of confirmatory reasoning, and participants in the low trust in feelings condition would experience an even lower level of confirmatory reasoning. Finally, participants in the control condition were asked to write their knowledge of the pyramids in Egypt, and what kind of traits are those of a good neighbour (Xu and Wyer 2012). Participants were given 3 minutes per each essay, although it is not strictly constrained as the web page would not proceed on its own. The full-length survey of the essay sections can be seen in Appendix A.

Product Evaluations

After the counterarguing manipulation, all participants were exposed to an evaluation of a

stepwise choice between two sports shoes (Chaxel and Han 2018). Following the method of

Chaxel and Han (2018) study, we labelled these two sports shoes with a neutral name, “Brand

J” and “Brand K”. These two products had three attributes, which are the sole unit, design

and colours, and online reviews. Right after assessing each product attribute, participants had

to answer two questions. The first question set was a diagnostic scale wherein participants

rate from 1 (information strongly favours Brand K) to 9 (information strongly favours Brand

J). This method had been pre-tested by Chaxel and Han (2018), which ensured that each

information given was neutral and did not favour either Brand J or K. The second question set asked participants about which brand was the leading option based on the information given to them (choice between J or K). Lastly, after participants had encountered all product attributes with the two sets of questions in between each attribute, they had to lay out their final decision. This final decision let participants choose between the two products as their final preference. In here, the preference bias for one option can be observed as participants progress towards each attribute and shows their final preference in the end (Meloy and Russo 2004).

The next section is a short series of three questions regarding participants’ general trust in feelings across various situations, and one attention check question. The first question asked the level of certainty that participants had upon making their final decision of shoe brand from 0 (absolutely uncertain) to 100 (absolutely certain) bar scale. This question was then followed up by an attention test to see if participants had been paying attention to the text or not, and to prevent mechanical bots from succeeding the test. Sometimes it is not obvious for researchers to identify these participants; but it is necessary to spot them as they reduce the power of the experiment (Oppenheimer, Meyvis, and Davidenko 2009). In this experiment, we implemented the Instructional Manipulation Check (IMC) by Oppenheimer et al. (2009).

This tool consists of a question that is similar to other questions provided in the survey; in our case, a Likert scale. It simply asked participants to choose the number 7 out of a 1-7 point scale to prove that they had been paying attention. If participants failed the IMC, it means that they also failed to follow other instructions in the survey. Eliminating these participants would, in turn, increase statistical power (Oppenheimer, Meyvis, and Davidenko 2009). In the third question, participants were asked whether they trusted their gut feelings when choosing the shoe brand from a 1 (Not at all) to 7 (A lot) point scale. The last question asked participants “to what extent did you based your choice on how right the brand felt?” (1 =

“not at all”; 7 = “a lot”). These two items were averaged into a “reliance on feelings score”

which served as a manipulation check. It is expected that participants in the high trust in

feelings condition would score higher on this reliance on feelings measure than participants

in the low trust in feelings condition. The full-length version of this part of the survey can be

seen in Appendix B.

Results

This section discusses the results of individuals’ trust in feeling tested with confirmation bias.

Furthermore, the result of the moderation effect will be elaborated based on the main effect.

Manipulation check

A manipulation check is conducted beforehand in order to observe if the priming of high trust in feelings (HTF) condition or low trust in feelings (LTF) condition were successful. A variable named “TiF” (Trust in Feeling) was created with a nominal value (HTF = 1, LTF = 0). Another variable named “AVGTiF” (Average trust in feelings) was formed through averaging the last two 7-point scale questions regarding the reliance of feeling that served as our manipulation check. An independent-sample t-test was conducted to observe the main effect. We would like to know whether participants that have an HTF condition versus LTF condition differ in terms of their reliance on feelings. At the confidence interval of 95%, we found no significant differences of the main effect between high trust in feelings participants (M

HTF

= 5.18, SD

HTF

= 1.36), and low trust in feelings participants (M

LTF

= 5.28, SD

LTF

= 1.35); (t(277) = .61, p = .55, d = .07). This means that participants with an HTF condition LTF condition did not differ in terms of their reliance on feelings. Hence, our manipulation check has shown an unintended effect.

Furthermore, we would like to know if our independent variable and moderator variable are independent of each other. Therefore, we conducted a two-way ANOVA test in order to analyse orthogonality. Along with the “TiF” and “AVGTiF” variable, another variable called

“Counterarguing” was created to measure the moderating variable (Counterarguing = 1,

Control = 0). For example, if a participant were exposed to an HTF/Control condition, then

they would be labelled as TiF = 1, Control = 0. At the confidence interval of 95%, the

interaction effect of trust in feelings and counterarguing proved to be not significant, F(1,

275) = 2.06, p = .15, η

p2

= .007. This means that our independent variable and moderating

variable are independent of each other; thus, orthogonality does not apply to our model.

Orthogonality test result

F df p

ηp2

Manipulation check

Constant 3919.39 1 .00** 0.93

Trust in feelings conditions 0.56 1 0.45 0.01

Counterarguing 1.52 1 0.22 0.01

Participants' reported reliance on feelings

Trust in feelings conditions *

Counterarguing 2.06 1 0.15 0.01

Note: computed using  = .05, *p < .050, **p < .010, ***p < .001 Table 2: Orthogonality test result of a 2x2 ANOVA test

Main Analysis

We computed an average information distortion score (DVdistortion) for all of our participants combined. In this context, information distortion is calculated by the absolute scale on how participants rated their shoe favourability in accordance with Russo et al.

(1998)’s method. Out of a 1 to 9-point scale, with 5 being neutral, we recoded these variables with SPSS’s syntax by labelling a positive difference if the participant consistently favours the leading brand from one question to the other. On the other hand, a negative label will be branded if participant deviates from their previous choice to choose the opposing brand. For example, if a participant rated brand J as their favoured brand on the 2

^nd

attribute (design &

colour), and then rate the next information on the 3

^rd

attribute a 1 on a scale of 1 to 9 – favouring brand K – then the absolute score will be: -abs (1-5) = -4. The absolute score is coded negatively if participants initial choice (i.e. based on exposure to the first attribute) does not match their rating of the subsequent 2

^nd

attribute. In contrast, if a participant chose brand J as their leading choice, and then rating the information a 9 on a scale of 1 to 9 – favouring J – then the absolute score will be: abs (9-5) = 4. The absolute score is coded positively if participants initial choice matches their rating of the subsequent 2

^nd

attribute. It should be noted that we begin to measure information distortion from the 2

^nd

question of the 1

^st

attribute (design & colours), and not from the 1

^st

question. This is because, at the beginning of the 1

^st

attribute, participants did not have a preference yet.

Based on our descriptive statistics on information distortion, the average distortion for the

HTF/Counterarguing condition is smaller (M = .18, SD = 2.04) than the HTF/Control

condition (M = .43, SD = 2.31). This may imply that participants in the HTF/Counterarguing

condition are less biased in their product preference in comparison to participants in the

HTF/Control condition. Furthermore, the average distortion of the LTF/Counterarguing condition is a bit smaller (M = .10, SD = 2.25) than the LTF/Control condition (M = .14, SD

= 2.54). This may imply that participants in the LTF/Counterarguing condition are less biased in their product preference in comparison to participants in the LTF/Control condition.

However, we may not derive anything conclusive yet basing only from these results. A one- way ANOVA test is conducted to investigate the main effect of trust in feelings (TiF) on information distortion (DVdistortion). At the confidence interval of 95%, the main effect of trust in feelings on information distortion proves to be insignificant, F(1, 277) = .53, p = .47.

This means that the high trust in feelings condition and the low trust in feelings conditions did not differ significantly in terms of confirmatory bias. Thus, the first hypothesis (H1) is rejected, and the null hypothesis is accepted.

Furthermore, a Macro PROCESS Hayes Model 1 of moderation was carried out with bootstrapping samples of 10,000 and the confidence interval of 95%. We examined the interaction effect of trust in feelings and counterarguing on information distortion. The result shows an insignificant interaction effect of trust in feelings and counterarguing on information distortion, b = -.20, t(275) = -.36, p = .72. Afterwards, we examined the direct effect of trust in feelings on information distortion. The result shows an insignificant direct of trust in feelings on information distortion, b = .19, t(275) = .69, p = .49. Lastly, we examined the direct effect of counterarguing/control condition on information distortion. The result shows an insignificant direct effect of counterarguing/control condition on information distortion, b = -.16, t(275) = -.60, p = .55. This means that the different levels of trust in feelings primed with counterarguing or control conditions did not differ significantly in terms of confirmatory bias. Thus, the second hypothesis (H2) is rejected, and the null hypothesis is accepted.

In addition to that, we conducted a Macro PROCESS Hayes model 3 LOGIT (moderated moderation) in order to see the probability if the first choice of shoe brand would end up to as the final choice as well as a function of trust in feelings and counterarguing. As the dependent variable, we used the final choice outcome of Brand K versus Brand J. For the independent variable, we used the first choice of the shoe brand, and counterarguing as the moderator.

Since this is a moderated moderation, we also used TiF (trust in feelings) as a second

moderator. The result of this analysis provided us with insignificant interactions of all

variables except for Choice 1 (First choice) at b = 2.0068, x

²

(1,1) = 7.0257, p < .0001. Which

means that all participants are more likely to maintain their first preference until the final choice. Nevertheless, this result does not provide significant evidence on any of our hypotheses.

Figure 2: Three-way moderation

Discussion

The main aim of this research is to examine how trust in feelings affect information distortion in the consumer decision-making process. It is essential to take notes that our result should be analysed with caution. Our manipulation check yielded insignificant result and therefore, did not result in the initial intended manipulation. As shown by the main analysis section, we have found no statistical evidence to support our first hypothesis. According to our findings, there is no significant relationship between high trust in feelings versus low trust in feelings condition on confirmation bias. Furthermore, we do not find that our trust in feelings manipulation trigger participants in processing information more/less cognitively or feeling- based. This means that participants in the high trust in feelings condition and participants in the low trust in feelings condition does not differ in their level of information distortion.

However, we did find out that once a participant has a leading preference for a shoe brand, they are very likely to stick to their choice until the end, in line with Chaxel et al. (2018).

Nevertheless, this finding does not provide evidence for any of our hypotheses.

Aside from the independent variable, this research examined whether there is a significant

interaction effect of trust in feelings moderated by counterarguing on confirmation bias. Here,

we utilised Hayes’ PROCESS Model 1 to test whether counterarguing/control condition

moderates the main effect of trust in feelings on confirmation bias. Similar to the first hypothesis, we have found no evidence that supported our second hypothesis. According to our result, we found out that counterarguing mindset did not moderate the relationship between trust in feelings and confirmation bias. High trust in feelings individuals primed with a counterarguing mindset do not experience a lower level of confirmatory reasoning, and low trust in feelings individuals primed with counterarguing mindset will not experience an even a lower level of confirmatory reasoning. However, we already expected that our second hypothesis would not be significant as it is dependent on our first hypothesis. Furthermore, we find no evidence that counterarguing produced any difference in the level of information distortion. These findings are in contrast to Chaxel et al. 2018, which means that neither the counterarguing condition nor the control condition trigger participants in processing information subjectively or objectively. As we have mentioned earlier, we have found out that once a participant has a leading preference for a shoe brand, there is a high probability that they will stick with their choice until the end. However, we also did not find any evidence that counterarguing of consistency mindset mitigated information distortion of participants.

Limitations

The first limitation of this research could be attributed to the fact that it was conducted via Amazon Mechanical Turk (MTurk), which is an online platform where workers from all around the world could fill up surveys for money. We made a filter of only workers who are experienced in 5000 or more surveys and has a 97% approval rate that can access and do our survey. However, some MTurk participants were seasoned workers, and have been exposed to surveys such as this and knew the real context of the experiment as they have encountered this type of study several times before our study (Chandler, Mueller, and Paolacci 2014). This is a dilemma because it is important to filter our participants based on their experience (over than 1,000 Human Intelligence Tasks (HITs)). If this is the case, they might also already know what to write in the essay sections, and what to choose in the multiple-choice section.

Hence, our manipulation would not work with these people as they are aware of the real nature of this survey.

The second limitation revolves around the time taken by MTurk workers to finish the survey.

According to our descriptive statistics, the average time taken by participants to finish the

survey is 6.8 hours (M

time

= 24,486.4588 seconds), with the fastest being 12.28 minutes (Min

time

= 737 seconds), and the longest time taken of 11 days (Max

time

= 960960 seconds).

Within 6.8 hours, participants might leave the study and come back another time to do it.

This might as well wear off any priming that had occurred towards participants; and thus, could be the reason why our manipulation yields insignificant result. Moreover, leaving the study meant that participants might forget, for example, the attributes of shoes J and K, or even the examples given to answer the essay sections. Several possible reasons for this behaviour could be because this survey takes quite a long time to fill in, and it is lengthy in terms of question numbers. The 1

^st

set of essays had a 4 minutes timer (HTF/LTF), the 2

^nd

set of essays had a 3 minutes timer (counterarguing/control), followed by six multiple-choice questions for confirmation bias, and three multiple-choice questions regarding trust in feelings. In addition to that, according to Necka et al. (2016), MTurk participants are more likely to multitask and leave studies while they are doing it in comparison to participants that are not crowdsourced. If this is the case, answers provided by participants will then be unreliable because they had forgotten about the information that could be crucial to help them understand the context better. Overall, MTurk participants are prone to exhibiting behaviour that would be problematic in an experiment (Necka et al. 2016).

The third limitation is connected to the number of participants that we have filtered out. Out of 521 participants, we excluded 242 due to the criteria mentioned earlier in the methodology section above. However, one criterion stood up as the most time-consuming to observe:

quality of the essay section. It was important that each essay is analysed for its authenticity, fluency in English, and whether or not it made sense in accordance to the question asked. The 242 participants that we excluded had a similar pattern in answering the essay section: literal copy-pasted sentences with full citation of sources, same answers between participants, or answers that were seemingly too long and eloquent to be true despite the time pressure. It has been studied that MTurk participants are more likely to use search engines or ask information from other participants in order to obtain answers for studies than traditional participants (Necka et al. 2016). Even after the effort of excluding 242 participants, we have no way of knowing whether the remaining 279 participants were truthful with their way of answering or not.

The fourth limitation is individual differences in trusting their feelings. Some inherently rely

lead to the right direction. Other individuals, on the other side, might not rely so much upon their feelings and perceive it as unreliable/untrustworthy. As Avnet et al. (2012) have stated, an individual’s perceived history in trusting their feelings and social-cultural environment may be the cause of these individual differences in reliance on feelings. Some might experience historically successful reliance on feelings, and some might not. Some culture might encourage the usage of feelings for decision-making processes, and some might encourage cognitive analysis instead. In this case, manipulating trust in feelings would vary in results depending on how much do participants rely on their feelings.

Implication

Despite all of the limitations, there are several implications that might be important for marketers and relevant researchers and individual in the field of marketing and business. We believe that our research model is novel, as prior research has shown that trust in feelings activates consistency mindset (Lee, Amir, and Ariely 2009), and that consistency mindset whether it is bolstered or counterargued result in a lower or higher level of confirmation bias (Chaxel and Han 2018). Nevertheless, there is no research available regarding the effect of trust in feelings towards confirmation bias. With this information in hand, we believe that our study could be utilised as a platform for researchers in the marketing field that are eager to analyse the relationship of consumer psychology and behaviour towards purchase decisions.

As our research has failed to see any relationships between trust in feelings towards

confirmation bias/information distortion, new questions arise that might be interesting to be

studied. Namely, if this experiment can only be successful through lab/traditional surveys,

how would marketers be able to implement the “priming” of high and low trust in feelings in

real life towards customers? Alternatively, is it an inherent characteristic that is hard to

manipulate in real life? Although we have failed to identify factors that could significantly

mitigate information distortion, we suggest researchers and marketers seek further ways to

eliminate or reduce information distortion through, e.g. advertisement, or other methods that

may trigger the cognitive analysis instead of the affective analysis of consumers.

Future Research

We suggest future research to observe the relationship between trust in feelings with confirmatory reasoning, moderated by counterarguing mindset in an experimental rather than an online survey setting. Monitoring participants on the spot might yield a more reliable and valid result. A controlled setting will provide researchers with the ability to ward off externalities that may interrupt the flow and result of the experiment. Even though we have filtered out many participants due to their ineligibility according to our standards, there might still be some participants that were mechanical bots; it is hard to differentiate their answers from real human beings.

Previous research has shown that trust in one’s own feelings activates consistency mindset (Lee, Amir, and Ariely 2009). Furthermore, the existence of a consistency mindset in individual results in confirmation bias upon information processing (Chaxel and Han 2018).

There has to be a correlation between these three variables, especially linked by consistency mindset. It might be interesting to implement another moderation experiment or a mediation of this model; preferably in a physical experimental setting rather than in an online platform.

Future research could also explore the connection between trust in feelings and confirmation bias with a different level of consistency mindset such as the bolstering/boosting of consistency mindset. It could be further studied whether or not the act of bolstering will enhance or weaken information distortion. In conclusion, this interaction of trust and feelings in relation to confirmation bias, moderated by counterarguing (consistency mindset) have not been studied before. This study could start as a stepping stone for other researchers to explore further possibilities regarding this relationship and other seemingly related variables.

Conclusion

This research has examined how different levels of trust in feelings within an individual

affect information distortion in decision-making. Furthermore, this research also examined

the moderating effect of counterarguing mindset. Based on online surveys of 279 participants

from various nationalities from Amazon Mechanical Turk, we found out that our result does

not correspond with our proposed hypotheses. In summary, information distortion levels in

the decision-making process are not affected by the level of trust in feelings within

individuals. Moreover, with differing levels of trust in feelings, the level of information

distortion is not increased or decreased with counterarguing mindset.

References

Avnet, Tamar, Michel Tuan Pham, and Andrew T. Stephen (2012), “Consumers’ Trust in Feelings as Information,” Journal of Consumer Research, 39 (4), 720–35.

Beauvois, Jean-Léon, Robert-Vincent Joule, and Robert Joule (1996), A radical dissonance theory, Taylor & Francis.

Beckmann, Jürgen and Julius Kuhl (1984), “Altering information to gain action control:

Functional aspects of human information processing in decision making,” Journal of Research in Personality, 18 (2), 224–37.

Blumberg, Stephen J., Daniel T. Gilbert, Elizabeth C. Pinel, Thalia P. Wheatley, and Timothy D. Wilson (1998), “Immune neglect: A source of durability bias in affective

forecasting,” Journal of Personality and Social Psychology, 75 (3), 617–38.

Brownstein, Aaron L, Stephen J Read, and Dan Simon (2004), “Bias at the racetrack: Effects of individual expertise and task importance on predecision reevaluation of alternatives,”

Personality and Social Psychology Bulletin, 30 (7), 891–904.

Carlson, Kurt A and J Edward Russo (2001), “Biased Interpretation of Evidence by Mock Jurors,” Journal of Experimental Psychology: Applied, 7 (2), 91–103.

Chandler, Jesse, Pam Mueller, and Gabriele Paolacci (2014), “Nonnaïveté among Amazon Mechanical Turk workers: Consequences and solutions for behavioral researchers,”

Behavior research methods, 46 (1), 112–30.

Chaxel, Anne Sophie and Yegyu Han (2018), “Benefiting from disagreement:

Counterarguing reduces prechoice bias in information evaluation,” Journal of Consumer Psychology, 28 (1), 115–29.

———, J. Edward Russo, and Catherine Wiggins (2016), “A Goal-Priming Approach to Cognitive Consistency: Applications to Judgment,” Journal of Behavioral Decision Making, 29 (1), 37–51.

Denes-Raj, Veronika and Seymour Epstein (1994), “Conflict between intuitive and rational processing: When people behave against their better judgment.,” Journal of personality and social psychology, 66 (5), 819.

Epstein, Seymour (2003), “Cognitive‐experiential self‐theory of personality,” Handbook of psychology, 5, 159–84.

Festinger, Leon (1962), A theory of cognitive dissonance, Stanford university press.

Gawronski, Bertram and Fritz Strack (2012), Cognitive consistency: A fundamental principle

in social cognition, Guilford press.

Hart, William, Dolores Albarracın, Alice H Eagly, Inge Brechan, Matthew J Lindberg, and Lisa Merrill (2009), “Feeling Validated Versus Being Correct : A Meta-Analysis of Selective Exposure to Information,” Psychological Bulletin, 135 (4), 555–88.

Jonas, Eva, Stefan Schulz-Hardt, Dieter Frey, and Norman Thelen (2001), “Confirmation bias in sequential information search after preliminary decisions: an expansion of dissonance theoretical research on selective exposure to information.,” Journal of personality and social psychology, 80 (4), 557.

Kostopoulou, Olga, J Edward Russo, Greg Keenan, Brendan C Delaney, and Abdel Douiri (2012), “Information Distortion in Physicians’ Diagnostic Judgments,” Medical Decision Making, 32 (6), 831–39.

Kruglanski, Arie W and Garriy Shteynberg (2012), “Cognitive consistency as means to an end: How subjective logic affords knowledge.,” Cognitive consistency: A fundamental principle in social cognition, 245–64.

Lee, Leonard, On Amir, and Dan Ariely (2009), “In Search of Homo Economicus: Cognitive Noise and the Role of Emotion in Preference Consistency,” Journal of Consumer Research, 36 (2), 173–87.

Lingle, John H. and Thomas M. Ostrom (1979), “Retrieval selectivity in memory-based impression judgments,” Journal of Personality and Social Psychology, 37 (2), 180–94.

Loewenstein, George (1996), “Out of Control: Visceral Influences on Behavior,”

Organizational Behavior and Human Decision Processes, 65 (3), 272–92.

Lynch, Jr., John G., Howard Marmorstein, and Michael F. Weigold (1988), “Choices from Sets Including Remembered Brands: Use of Recalled Attributes and Prior Overall Evaluations,” Journal of Consumer Research, 15 (2), 169–84.

Meloy, Margaret G and J Edward Russo (2004), “Binary choice under instructions to select versus reject,” Organizational Behavior and Human Decision Processes, 93 (2), 114–

28.

Necka, Elizabeth A, Stephanie Cacioppo, Greg J Norman, and John T Cacioppo (2016),

“Measuring the prevalence of problematic respondent behaviors among MTurk, campus, and community participants,” PloS one, 11 (6), 19.

Oppenheimer, Daniel M, Tom Meyvis, and Nicolas Davidenko (2009), “Journal of Experimental Social Psychology Instructional manipulation checks : Detecting

satisficing to increase statistical power,” Journal of Experimental Social Psychology, 45 (4), 867–72.

Pham, Michel T (1998), “Use of Feelings in Decision Making,” Journal of Consumer

Research, 25 (2), 144–59.

Pham, Michel Tuan (2004), “The Logic of Feeling,” Journal of Consumer Psychology, 14 (4), 360–69.

Russo, J. Edward, Margaret G. Meloy, and Victoria Husted Medvec (1998), “Predecisional Distortion of Product Information,” Journal of Marketing Research, 35 (4), 438.

Russo, J E and Kevyn Yong (2011), “The distortion of information to support an emerging evaluation of risk,” Journal of Econometrics, 162 (1), 132–39.

Russo, J Edward, Kurt A Carlson, Margaret G Meloy, and Kevyn Yong (2008), “The Goal of Consistency as a Cause of Information Distortion,” Journal of Experimental

Psychology: General, 137 (3), 456–70.

———, Victoria Husted Medvec, and Margaret G Meloy (1996), “The distortion of

information during decisions,” Organizational behavior and human decision processes, 66 (1), 102–10.

———, Margaret G Meloy, and T Jeffrey Wilks (2000), “Predecisional distortion of information by auditors and salespersons,” Management Science, 46 (1), 13–27.

Schwarz, Norbert, Herbert Bless, and Gerd Bohner (1991), “Mood and persuasion: Affective states influence the processing of persuasive communications,” Advances in

Experimental Social Psychology, 24 (C), 161–99.

——— and Gerald L. Clore (1983), “Mood, Misattribution, and Judgements of Well-Being:

Informative and Directive Functions of Affective States,” Journal of Personality and Social Psychology, 45 (3), 513–23.

Siemer, Matthias and Rainer Reisenzein (1998), “Effects of Mood on Evaluative Judgements:

Influence of Reduced Processing Capacity and Mood Salience,” Cognition and Emotion, 12 (6), 783–805.

Snell, J., Gibbs, B.J. & Varey, C. (1995) (1995), “Intuitive Hedonics: Consumer Beliefs About the Dynamics of Liking,” Journal of Consumer Psychology, 4 (1), 33–60.

Svenson, Ola (1996), “Decision making and the search for fundamental psychological

regularities: What can be learned from a process perspective?,” Organizational behavior and human decision processes, 65 (3), 252–67.

Tyszka, Tadeusz and Marek Wielochowski (1991), “Must boxing verdicts be biased?,”

Journal of Behavioral Decision Making, 4 (4), 283–95.

White, Katherine and Cathy McFarland (2009), “When are moods most likely to influence

consumers’ product preferences? The role of mood focus and perceived relevance of

White, Victoria M, Alexander J Wearing, and David J Hill (1994), “Is the conflict model of decision making applicable to the decision to be screened for cervical cancer? A field study,” Journal of Behavioral Decision Making, 7 (1), 57–72.

Wright, Peter (1975), “Consumer Choice Strategies: Simplifying vs. Optimizing,” Journal of Marketing Research, 12 (1), 60–67.

Xu, Alison Jing and Robert S. Wyer (2012), “The Role of Bolstering and Counterarguing

Mind-Sets in Persuasion,” Journal of Consumer Research, 38 (5), 920–32.