Halo Effects in Advice-Seeking Decisions

This is a lab report I wrote for one of my modules on epistemic spillovers and the halo effect that got a first.

Word Count: 3255

Abstract

Our study investigates the presence of the halo effect on advice-seeking behaviours in the presence of competence and warmth. 110 participants partook in a within-subjects experimental design. Their perceptions of competence and warmth were manipulated through learning information about accuracy on general knowledge quizzes and generosity charitable-giving respectively. Participants rated sources for competence and decided which source they would seek advice from when answering general knowledge questions (‘source-choice’). Results demonstrated that accurate sources were rated as more competent and chosen significantly more to provide advice. We did not find positive evaluations of generosity to extend towards positive competence evaluations or an increase in advice-seeking behaviour. We were unable to uncover the presence of a halo effect in the presence of warmth on advice-seeking behaviour but manage to identify several limitations for the apparent lack of effect and indicate pathways for future research.

1. Introduction

When faced with important decisions, people feel anxious, a state of physiological arousal in response to potential undesirable outcomes (Brooks & Schweitzer, 2011). Because anxiety impairs information processing, we actively seek advice from others (Eysenck, 1992). This can create conflict when our internal judgement and external advice differ, heightening anxiety and amplifying subsequent advice-seeking behaviours in a positive feedback cycle (Yaniv & Kleinberger, 2000). Anxious individuals are less able to discriminate between good and bad advice, a fundamental skill required in everyday life (Gino et al., 2012). Elucidating mechanisms behind advice-seeking decisions are essential to alleviate anxiety for rational, favorable behaviour.

We seek advice from those we consider the most accurate or similar to ourselves (Marks et al., 2018). This may result from social desirability bias, where we follow the advice of similar others to maintain connectedness to our in-group (Byrne, 1969; Jiang et al., 2010). Alternatively, preferential advice-seeking may be due to our implicit belief that similar others hold preferences more aligned to our own, making their advice seem more accurate and diagnostic (Del Vicario et al., 2016). Boorman et al. (2013) found that participants credited others when their opinions were similar, but penalised them when opinions conflicted. Our bias towards similar others interferes with our objective and accurate assessment of their skills, leading us to believe they are more competent (Aaker et al., 2010). This creates an illusion of understanding of our advisor’s mental state, inducing a sense of certainty that validates the advice, increasing its persuasive value (Faraji-Rad et al., 2015). We thus seek advice from similar others, creating a feedback loop with real-world consequences for the spread of fake news and political polarisation (Duckerman et al., 2013).

Marks et al. (2018) highlight that epistemic spillovers occur when similarity in one field generalises to biased assessments in others. Shared political convictions influenced participant’s desire to consult and use other’s (‘sources’) views in domains without political relevance. This occurred even when participants explicitly observed that sources with differing political views were objectively more proficient in providing accurate answers. We attribute this to the halo effect, which is the tendency for an evaluation in one area to influence appraisal of another on the basis of a small sample of known positive or negative features (Nisbett & Wilson, 1977). Learning about their political similarity to a source interfered with participants’ processing. Generalising competence in one domain to another culminated in a halo effect which reduced participant’s objective assessment of sources, influencing their desire to consult and use source’s views even on tasks unrelated to politics. Our automatic preferences for those with similar convictions creates a positive affect which spills over into our evaluation of other, unrelated domains (Marks et al., 2018).

Kahneman & Frederick (2002) proposed that this is alternatively due to the presence of heuristics – mental shortcuts for making inferences without always utilising all available information in order for efficient, rapid information processing (Katsikopoulos, 2011). Heuristics operate concurrently with beliefs that like-minded people are more worth consulting, resulting in an extension of the contexts to which this belief applies regardless of its relevance, again alluding to the halo effect (Hovland et al., 1953).

While halo effects assess a person’s quality and personality to evaluate similarity, placebo effects evaluate symptoms for generalisation (Fabrizio, 2014). Both emphasise the role of context and social behaviour. Social behaviour moderates the effect of expectations on physiological outcomes, powerfully influencing placebo responses (Howe et al., 2017). Researchers investigated whether placebo effects would arise by altering social context through manipulating competence and warmth of doctors. Competence judgements included perceptions of effectiveness and capability, whereas warmth judgements connoted generosity, sincerity and trustworthiness (Judd et al., 2005). Doctors positively demonstrating both traits had greatest influence over placebo effects. This occurred because judgements of warmth and competence underlie our perceptions of others (Cuddy et al., 2008). Warmth culminates in illusions of trust and alliance, while competence represents one’s ability to achieve desired outcomes (Peeters, 1983).

However, Aaker et al. (2010) found that warmth only influenced advice-seeking behaviour in the presence of competence. Grandey et al. (2005) also found that isolated warmth-judgements would not translate to an increase in advice-seeking behaviour, whereas isolated competence-judgements would. Despite this conflicting literature, we propose that warmth is necessary to elicit the placebo effect as our advice-seeking decisions depend on our perceptions of the expertise and trustworthiness of the advisor (Hofmann et al., 2009; Marks et al., 2018; Howe et al., 2017).

What remains unclear is whether warmth would similarly influence advice-seeking decisions that occur due to the halo effect. As an extension of Marks et al. (2018), we posit that the extent to which a source is considered competent and warm will spillover into subsequent advice-seeking decision-making in participants because of the halo effect. We manipulate perceptions of competence and warmth through information about accuracy on general knowledge quizzes and charitable-giving respectively. Participants first learnt about source’s (i) warmth (operationalised by generosity in charitable giving) and (ii) competence (operationalised by accuracy in general knowledge questions). After rating sources on these characteristics, participants completed the second part of the study where they decided which source to turn to for advice when answering general knowledge questions. Participants were rewarded for accuracy and provided with an economic incentive as motivation to complete the task to the highest standard.

We hypothesised that positive evaluations of accuracy and generosity on general knowledge questions will (i) generate more positive perceptions of source competence and (ii) determine which sources participants chose to hear from when doing the general knowledge quiz. We hope to explicate whether perception of generosity interferes with ability to learn about another’s competency in unrelated general knowledge tasks where it is in participant’s best interest to learn who objectively excels at the task in order to turn to them for assistance. We expect positive evaluations of generosity and accuracy to extend to positive evaluations of competence, as well as increase advice-seeking behaviour towards generous, accurate sources.

2. Method

2.1 Participants

110 undergraduates from UCL participated (9 males, 42 females). Ages ranged from 18 to 22 (Mage = 18.8, SDage = 0.75). A £40 cash incentive was offered to the participant achieving the highest score on this study. All participants consented to the experiment after being debrief and ethical approval was obtained from UCL Psychology’s Ethics Committee.

2.2 Design

A 2x2 within-subjects design was used. Participants took part in all experimental conditions. There were two experimental sessions, (1) warmth and generosity and (2) political similarity, but only session (1) is included in this report.

Independent variables were perceived (a) accuracy of sources (operationalised through whether sources were correct or incorrect) and (b) generosity of sources (operationalised through whether sources donated to charity). This created four experimentally manipulated conditions: accurate-generous, accurate-ungenerous, inaccurate-generous and inaccurate-ungenerous. Two dependent variables were measured and analysed separately: (i) competence ratings (operationalised as mean competence ratings) and (ii) source choice (operationalised by the percentage of times each source was chosen).

A two-factor Analysis of Variance (ANOVA) test examined the influence of our independent variables on our dependent variables and to determine the presence of an interaction.

2.3 Materials and Stimuli

The questionnaire was created using Qualtrics.

2.3.1 Learning Stage

1. Charity trials. 40 charities from the top 100 UK charities according to the Charity League Table (Morar HPT, 2018) were presented to participants (1/trial).

2. General knowledge trials. 40 general knowledge questions from the hardest ‘Who Wants to be a Millionaire?’ questions (Shaw, 2019) were presented to participants (1/trial).

2.3.2 Choice Stage

24 general knowledge questions, derived in the same way described above, were presented to participants.

2.3.3 Sources

Sources were presented to participants as animal icons of a bird, an elephant, an angelfish and a leopard (Google Images).

2.4 Procedure

Experimental sessions were conducted on individual computers in a communal room in UCL over 60-minute sessions. Participants were provided with hyperlinks on desktops to access the questionnaire (Qualtrics).

2.4.1. Learning Stage

There were 8 blocks of 40 trials each (10 charity trials and 10 general knowledge trials interleaved). Qualtrics’ loop and merge tool randomised the order of questions in each block. Participants learnt about ‘other participant’s’ (hereafter ‘sources’) generosity to charity and competency on the general knowledge tasks.

1. Charity trials. Each trial proposed a hypothetical situation: sources were given £1. They could either give £0.50 (and keep £0.50) to charity or keep the full £1.00. A charity was presented and participants bet on which sources gave £0.50 and which kept £1.00 by clicking on any of the four icons.

2. General knowledge trials. Each trial presented a different general knowledge question to participants. Participants were shown the binary choice sources had. They then bet on which sources they believed answered the question accurately and inaccurately.

In both trials, participants entered their responses in their own time before being shown the source’s actual responses. Thereafter they were given feedback on whether their bets were accurate.

2.4.2. Sources

Participants were told that the source’s responses they were shown on each trial were gathered from data of participants who performed the task at an earlier stage. Participants were unaware that sources were algorithms designed to respond in the following pattern:

Source 1. Accurate-Generous – selected ‘give’ on 80% of charity trials; correct on 80% of general knowledge trials.

Source 2. Accurate-Ungenerous – selected ‘keep’ on 80% of charity trials; correct on 80% of general knowledge trials.

Source 3. Inaccurate-Generous – selected ‘give’ on 80% of charity trials; inaccurate on 50% of general knowledge trials.

Source 4. Inaccurate-Ungenerous – selected ‘keep’ on 80% of charity trials; inaccurate on 50% of general knowledge trials.

Pictures of animals were chosen to represent sources to avoid potential gender or racial bias. To reduce order effects, pictures assigned to sources were algorithmically randomised for counterbalancing.

2.4.3. Ratings Stage

Participants were presented with questions (e.g. ‘How competent was the source at answering general knowledge questions?’; ‘How generous was the source on charity rounds?’) which they rated individually on separate 6-point Likert Scales (1 = Very Incompetent/Ungenerous, 6 = Very Competent/Generous). Identical questionnaires were completed for each source.

2.4.4. Choice Stage

This stage assessed which source participants wanted to hear from about general knowledge questions and how they used the information they were shown. 4 trials were conducted for each of the 6 possible source pairs which were pseudo-randomised throughout. On each of the 24 trials, participants were presented with a new general knowledge question and were given a binary choice. They rated their confidence in their answer on a scale of 0 = ‘just guessing’ to 100 = ‘completely confident’. They were presented with two sources and chose one to reveal that sources’ response. Participants were given the chance to change their answer. Lastly, they rated their confidence in their final decision. All decisions were self-paced.

2.4.5. Attention Checks

Participants were asked several multiple-choice questions about the trial they had just completed as an attention check following both the learning and choice stages.

2.4.6. Post-Task Ratings and Debrief

Participants self-reported their perceptions of each source’s (1) competency on general knowledge questions and (2) generosity on a 6-point Likert Scale (1 = Very Incompetent/Ungenerous, 6 = Very Competent/Generous).

Participants were given the opportunity to withdraw their data.

3. Results

3.1. Perceptions of competence on general knowledge tasks were influenced by accuracy of sources rather than by generosity (charitable giving)

Participants rated accurate sources as more competent. A 2accuracy x 2generosity repeated-measures ANOVA revealed a main effect of source accuracy (F(1,50) = 23.13, p = .001, η2 = .46). There was no illusory main effect of generosity (F(1,50) = 0.11, p = .74, η2 = .002) or interaction effect (F(1,50) = 1.34, p = .25, η2 = .03), indicating that generosity did not impact participant’s perceptions of source competence but the accuracy of sources did.

Despite both accurate sources being correct in 80% of trials, participants rated accurate-generous sources (µ = 4.69, SD = 1.10) as marginally more competent than accurate-ungenerous sources (µ = 4.61, SD = 1.06). Likewise, both inaccurate sources were correct in 50% of trials, inaccurate-generous sources (µ = 3.69, SD = 0.99) were rated as marginally less competent than inaccurate-ungenerous sources (µ = 3.86, SD = 1.06).

Graph 1: Descriptive Statistics for Competence (1 = very incompetent; 6 = very competent)

3.2. The lack of an illusory perception of competence meant that the relationship between generosity (charitable giving) and information-seeking behaviour (source choice) was not mediated

Accurate sources were chosen significantly more than inaccurate sources (Graph 2). While accurate-generous sources (µ = 30.07, SD = 14.68) were chosen marginally more than accurate-ungenerous (µ = 28.51, SD = 13.19) sources, inaccurate-generous sources (µ = 19.53, SD = 14.03) were chosen marginally less than inaccurate-ungenerous sources (µ = 21.9, SD = 15.09).

Objective accuracy predicted likelihood of participants’ choosing the source – a 2accuracy x 2generosity repeated-measures ANOVA revealed that accurate sources were chosen more often (F(1,50) = 11.00, p = .002, η2 = .22). This effect was not reduced when generosity was controlled (F(1,50) = 0.029, p = .87, η2 = .001), indicating a lack of interaction effect between accuracy and generosity (F(1,50) = 1.12, p = .30, η2 = .022).

Graph 2: Descriptive Statistics for Source Choice (1 = very incompetent; 6 = very competent)

4. Discussion

We find that participant’s competence ratings were influenced by source-accuracy on general knowledge tasks but not by source-generosity. Further, source-choice was influenced by source-accuracy but not by source-generosity.

Accuracy significantly predicted both competence ratings and source-choice. Accurate sources were perceived as more competent and were more sought after for advice during the choice stage. Following Marks et al. (2018), we hypothesised that interactions between positive views curated by generosity and objective accuracy should culminate in illusory perceptions of overall competence. We expected both generous and accurate sources to have greater competence ratings independently, but accurate-generous sources to have highest competence ratings. Our rationalisation was that generosity should interfere with assessments of competence; we believed generosity would enhance perceptions of source-accuracy by curating positive views of sources which then generalise to the unrelated domain of competence through the halo effect (Nisbett & Wilson, 1977).

Warmth, operationalised by generosity in charitable giving, was expected to elicit a halo effect when operating alongside competence. People recognising warmth in charitable giving as indicative of competence should generalise this belief to unrelated domains, for instance where warmth predicts competence in general knowledge tasks (Aaker et al., 2010). However, generosity neither influenced participant’s ratings of competence, nor impacted source choice. No halo effect was present as positive evaluations of generosity did not extend to positive evaluations of competence or influence participant’s desire to consult and use source’s views on general knowledge tasks.

A possible explanation is that our operationalisation of ‘warmth’ by manipulating ‘generosity’ retained its fundamental associations which acted as confounds. Warmth and competence are implicitly used to form perceptions (Judd et al., 2005). Consumers perceive non-profit organisations as warmer but less competent than for-profit organisations (Aaker et al., 2010). Charities are non-profit organisations, which may have elicited implicit feelings of incompetence that outweighed positive feelings of warmth generated by a source’s generosity. This could have negated the halo effect in which positive views of source’s generosity were generalised to their competence, resulting in the lack of effect warmth had on competence ratings and source choice (Aaker et al., 2010). However, a halo effect may be present in another respect. Implicitly associating non-profit organisations with warmth and incompetence could extend to sources: more charitable sources may have been perceived as warmer but less competent (Hansman, 1981). Future research could include an Implicit Associations Test to determine whether these biases are present and elucidate their impacted on our results.

Our manipulated variables differ in part from Marks et al. (2018)’s. First, ours was a culturally-diverse population sample of university students attending UCL rather than participants from the USA. The disparity in results could result from cultural differences – the diversity of our sample, with both collectivist (e.g. Asian cultures; China, Singapore) and individualist (e.g. Western cultures; Britain, USA) participants, differed from the solely individualist culture of Marks et al. (2018)’s sample. The collectivism of our Asian participants may have resulted in an increase in advice-seeking behaviour in the presence of generosity because warmth is a feature that maintains group harmony (Bond et al., 1982). Collectivist cultures find opinions and advice of group members important, increasing advice-seeking behaviours (Pavlou & Chai, 2002). This shaped participant’s eventual decisions to alter answers based on source choice (Kahled et al., 2006). Emphasis on group opinions is not as apparent in individualist cultures (Aaker & Maheswaran, 1997). The collectivism of our Asian participants may have mediated the effects of the individualism of our Western participants, resulting in a lack of effect as the opposing cultural values balance each other out. We should re-conduct our study using a sample of participants from the USA to remove this cultural confound.

Second, we presented the outcomes of the four sources jointly instead of linearly as Marks et al. (2018) had done in order to reduce task-length. Although this may have increased cognitive difficulty, we did this to lessen the fatigue and attention span required from participants. This ensured that participants maintained optimal focus throughout the task and mitigated the confound of their attention tapering off towards the end of a lengthy experiment, which was a limitation of Marks et al. (2018). However, this may have restricted participant’s holistic comprehension of source’s generosity – increasing cognitive difficulty can result in incomplete cognitive control and cognitive feedback (Hammond & Summers, 1972). Future research should present sources linearly – alleviating cognitive difficulty may result in a more significant generosity effects.

Finally, our lab study may have lacked ecological validity (Berkowitz & Donnerstein, 1982). Gino et al. (2012) found that anxious individuals sought and relied on advice from others more than those in neutral emotional states. Our study does not represent real-world decision-making as the effect of anxiety is likely to be lessened, or even ameliorated, by no real ramifications when making advice-seeking decisions. Important decisions, like choosing a new job, hold real consequences that are likely accompanied by high anxiety levels which correlate to an inability to discriminate between good and bad advice (Gino et al., 2012). This highlights the importance of understanding advice-seeking behaviours in order to alleviate anxiety and ensure rational, favorable advice-seeking behaviours prevail.

5. Conclusion

We found that accuracy was the core determinant of subsequent competence ratings and advice-seeking decision-making. We were unable to elucidate the presence of a halo effect or support the perception that generosity would influence advice-seeking behaviour. When looking at whether positive evaluations of generosity would extend to positive evaluations of competence and source choice, our results deviated from Marks et al. (2018)’s research on the influence of the halo effect on political similarity and decision-making. Perceptions of warmth did not manifest in halo effects from positive spillovers into evaluation of competence on general knowledge tasks. The contribution of our work can be qualified in our identification of several possibilities, such as cultural differences, ecological validity and cognitive difficulties, for the lack of effect which can be explored in future research into the halo effect on advice-seeking behaviour. This study further documents the influence of anxiety on decision-making and highlights the need to understand the adverse consequences this may have.

6. References

Aaker, J. & Maheswaran, D. (1997). The Effect of Cultural Orientation on Persuasion. Journal of Consumer Research. 24(3), 315 – 328.

Aaker, J. Vohs, K.D. & Mogilner, C. (2010). Nonprofits are seen as warm and for-profits as competent: Firm stereotypes matter. Journal of Consumer Research, 37(2), 224-237.

Berkowitz, L., & Donnerstein, E. (1982). External validity is more than skin deep: Some answers to criticisms of laboratory experiments. American psychologist, 37(3), 245.

Bogart, L. M., Bird, S. T., Walt, L. C., Delahanty, D. L., & Figler, J. L. (2004). Association of stereotypes about physicians to health care satisfaction, help-seeking behavior, and adherence to treatment. Social science & medicine, 58(6), 1049-1058.

Bond, M. H., Leung, K., & Wan, K. C. (1982). How does cultural collectivism operate? The impact of task and maintenance contributions on reward distribution. Journal of Cross-Cultural Psychology, 13(2), 186-200.

Boorman, E.D., O’Doherty, J.P., Adolphs, R, & Rangel, A. (2013). The behavioural and neural mechanisms underlying the tracking of expertise. Neuron, 80(6), 1558-1571. https://doi.org/10.1016/j.neuron.2013.10.024.

Brooks, A. W., & Schweitzer, M. (2011). Can nervous Nelly negotiate? How anxiety causes negotiators to make low first offers, exit early, and earn less profit. Organizational Behavior and Human Decision Processes. 115, 43–54. https://doi:10.1016/j.obhdp.2011.01.008.

Bryne, D. (1969). Attitudes and Attraction. Advances in Experimental Social Psychology. 35-89. New York: Elsevier.

Cuddy, A., Fiske, S. T., Glick, P., and Xu, J. (2002). A Model of (Often Mixed) Stereotype Content: Competence and Warmth Respectively Follow from Perceived Status and Competition. Journal of Personality and Social Psychology, 82(6), 878–902.

Cuddy, A., Fiske, S. T., & Glick, P. (2008). Warmth and competence as universal dimensions of social perception: The stereotype content model and the BIAS map. Advances in experimental social psychology. 40, 61–137.

Del Vicario, M., Bessi, A., Zollo, F., Petroni, F., Scala, A., Caldarelli, G., & Quattrociocchi, W. (2016). The spreading of misinformation online. Proceedings of the National Academy of Sciences, 113(3), 554-559. https://doi.org/10.1037/pnas.1517441113.

Duckerman, J.N., Peterson, E. & Slothuus, R. (2013). How elite partisan polarisation affects public opinion formation. American Political Science Review, 107(1), 57-79. https://doi.org/10.1017/S0003055412000500.

Eysenck, M. W. (1992). Anxiety: The cognitive perspective.

Fabrizio, B. (2014). Halo effects. Placebo Effects.

Faraji-Rad, A., Samuelsen, B. M., & Warlop, L. (2015). On the persuariveness of similar others: The role of mentalising and the feeling of certainty. Journal of Consumer Research, 42(3), 458-471. https://doi.org/10.1093/jcr/ucv032.

Gino, F., Brooks, A.W. & Schweitzer, M.E. (2012). Anxiety, advice, and the ability to discern: Feeling anxious motivates individuals to seek and use advice. Journal of Personality and Social Psychology, 102(3), 497-512. https://doi.org/10.1037/a0026413.

Grandey, Alicia, Glenda Fisk, Anna Mattila, Karen Jansen, and Lori Sideman (2005). Is ‘Service with a Smile’ Enough? Authenticity of Positive Displays during Service Encounters. Organizational Behavior and Human Decision Processes. 96(1), 38–55.

Hammond, K. R., & Summers, D. A. (1972). Cognitive control. Psychological review, 79(1), 58.

Hansmann, Henry B. (1981). Reforming Nonprofit Corporation Law. 129 (1), 497–623.

Hofmann, D. A., Lei, Z., & Grant, A. M. (2009). Seeking help in the shadow of doubt: The sensemaking processes underlying how nurses decide whom to ask for advice. Journal of Applied Psychology, 94(5), 1261–1274. https://doi.org/10.1037/a0016557

Hovland, C. I., Janis, I. L., & Kelley, H. H. (1953). Communication and persuasion.

Jiang, L., Hoegg, J., Dahl, D. W., & Chattopadhyay, A. (2010). The persuasive role of incidental similarity on attitudes and purchase intentions in a sales context. Journal of Consumer Research, 36(5), 778-791.

Judd, Charles M., Laurie James-Hawkins, Vincent Yzerbyt, and Yoshihisa Kashima (2005). Fundamental Dimensions of Social Judgment: Understanding the Relations between Judgments of Competence and Warmth. Journal of Personality and Social Psychology, 89(12), 899–913.

Katsikopoulos, K. V. (2011). Psychological heuristics for making inferences: Definition, performance, and the emerging theory and practice. Decision Analysis, 8(1), 10-29.

Kahneman, D., & Frederick, S. (2002). Representativeness revisited: Attribute substitution in intuitive judgement. Heuristics and biases: The psychology of intuitive judgement. 49-81.

Khaled, R., Biddle, R., Noble, J., Barr, P., & Fischer, R. (2006). Persuasive interaction for collectivist cultures. In Proceedings of the 7th Australasian User interface conference-Volume 50, 73-80.

Morar HPT. (2018). The 100 Most Valuable Charity Brands 2018: League Table. https://fundraising.co.uk/2018/09/26/morar-hpi/

Pavlou, P. & Chai, L. (2002). What Drives Electronic Commerce Across Cultures? A Cross-Cultueral Empirical Investigation of the Theory of Planned Behaviour. Journal of Electronic Commerce Research 3(4), 240 – 253.

Peeters, G. (1983). Relational and informational pattern in social cognition. Current Issues in European Social Psychology. 201–237.

Shaw, G. (2019). Could you answer these million-dollar questions from ‘who wants to be a millionaire?’. https://www.insider.com/hardest-questions-who-wants-to-be-a-millionaire-2018-12

Yaniv, I., & Kleinberger, E. (2000). Advice taking in decision making: Egocentric discounting and reputation formation. Organizational Behavior and Human Decision Processes, 83, 260 –281. https://doi:10.1006/ obhd.2000.2909.