Do We Have Stable Mental States, Such As Beliefs And Desires? Or Do We Invent These Moment-by-moment?

Do we have stable underlying beliefs and desires? Or do we invent them each moment? Do we have the mental depths implied by the former, or do we make up our minds as we go along? If the latter, then, mental depth is an illusion, and as well the notion of consistent, fixed preferences which guide the behaviour of the homo economicus . We shall review evidence suggesting preferences are invented moment-by-moment, but suggest that it is simplistic to argue for this view, or that of stable preferences. Reconciling these two positions is the challenge, but it first requires a richer understanding of what preferences even are. To this end, we will suggest the use of introspection with brain imaging methods.

Firstly, what is the case for moment-to-moment construction of preferences? We will follow Ariely and Norton s (2008, p.14) framework for determinants of behaviour and briefly review each component that offers evidence of on the spot decision-making:

Figure SEQ Figure * ARABIC 1. Determinants of behaviour

If preferences are stable and well-defined, then emotions should not influence choice. Yet, in Crockett et al. (2013), manipulating the level of serotonin which is involved in emotional regulation and empathy, influenced moral judgements and whether participants rejected unfair offers in the Ultimatum Game. Dutton and Aron (1974) showed how men having been taken across an unsteady bridge could misattribute adrenaline as arousal for the female interviewer. Even judgements of life satisfaction have been shown to be affected by mood, where mood was induced by interviewing participants on sunny or rainy days (Schwarz & Clore, 1983). Emotions are not stable, and clearly influence behaviour.

What about context? Economics assumes decision makers have complete preference orderings for all options, and will always choose that option highest in their orderings. Yet preferences have been shown to depend on the presence or absence of other options. Introducing a middle option in a choice set with two extremes makes it likelier to be chosen than if the middle option had appeared alone with either of the extremes. Adding an inferior option that is dominated by an existing option but not the other, will increase the probability of the dominating option being chosen. Finally, adding a competitive option similar to an existing option increases the probability of the dissimilar option being chosen. Context effects suggest the value of an option is temporarily affected by the other options the decision maker faces (Tversky & Simonson, 1993). They have been taken as evidence that a choice is reached by comparing available options. That means preferences are made up on the spot.

Environmental cues are like reference points, which subtly carry suggestions of what the appropriate, reasonable or normal thing to do is (Wansink & van Ittersum, 2013). In countries where people are required to opt-out of organ donation, there are significantly higher consent rates than those with an opt-in policy (Johnson & Goldstein, 2003). Can ticking a box, be as costly as to change preferences for organ donation? Wansink and van Ittersum (2013) showed that visual cues, such as portion size, will lead to substantially more consumption. Also a challenge for stable preferences are behaviours that seem to spread through social networks, such as obesity (Christakis & Fowler, 2007), smoking (Christakis & Fowler, 2008) and binge drinking (Ormerod & Wiltshire, 2009). If decision makers are guided by impulse, habit, and social norms, as environmental cues seem to suggest, then it casts doubt on the notion that beliefs and desires are stable. This summarises what Ariely and Norton (2008) refer to as situational factors.

Ariely and Norton (2008) also make the case for behaviour shaping preferences. They imagine people reflecting on their own behaviour, and from this, inferring what their preferences actually are. Thus choice itself, creates preference. They cite a number of studies to support this claim. Johansson, Hall, Sikstr m, and Olsson (2005) gave participants the task of choosing the most attractive female face from a pair of black-and-white pictures. Immediately after the choice, they were asked to describe the reasons for choosing as they did. Unknown to the participants, a card trick performed by the experimenter meant the two pictures had been switched. Only around 30% of participants noticed this change. This Choice Blindness (CB) methodology has been applied to other domains, such as political preferences. Hall et al. (2013) presented participants with a survey exploring the issues that a right-wing and left-wing coalition were divided on in a general election in Sweden. Using a magic trick, the experimenters altered some of the answers participants gave so that it placed them in the opposite political camp. Then, participants were asked to reason the attitudes they had expressed on these manipulated survey answers. Finally, each participant s political survey score was calculated. As in Johansson et al. (2005), few participants had noticed their replies had been manipulated (22%), and an overwhelming majority (92%) accepted their political survey score. If stable preferences exist, how could individuals endorse the opposite of the preference they had given a moment ago? It seems they could not even recollect their own views. In which case, peering inside mental depths does not seem possible since it relies on individuals that can remember themselves.

How can these results be explained? Cognitive dissonance (CD) makes the argument that individuals experience mental discomfort from holding two (or more) contradictory views (Harmon-Jones & Mills, 1999). For example, the political survey score was a dissonant cognition because it contradicted what participants thought about their political views. Human beings strive for internal consistency, and when dissonance is experienced, individuals feel discomfort, and are motivated to reduce this dissonance. In this case, by adopting the political survey score presented. CD can explain this making-up of preferences, moment-to-moment. Alternatively, according to self-perception theory (SPT) people develop their preferences, where they may have no previous experience, by observing their own behaviour and thus concluding what preferences must have caused it (Bem, 1972). That is, people interpret their own behaviour, in the same way as they would explain another s behaviour. In the case of Hall et al. (2013), individuals were unsure of their own political views. They observed what they thought was their own behaviour, and concluded that they are (apparently) closer to the views of the opposing political coalition.

Which of these theories are correct? SPT and CD are notoriously difficult to isolate as they make the same predictions (Johansson, Hall, T rning, Sikstr m & Chater, 2014). For our paper, they are only important in describing another channel by-which preferences seem to be made up on the spot. How important a channel it is, is rather dubious. Petitmengin, Remillieux, Cahour and Carter-Thomas (2013) replicated the Johansson et al. (2005) experiment with women s faces. Instead of putting individuals on the spot by having them justify their choice after being presented with the fact of their apparent decision, they used the interview elicitation method. This method of introspection involves prodding questions intended to get the decision maker to revisit the memories at the time of their choice. Using experimenter-enabled reflection, lasting 17 to 45 minutes, resulted in detection rates of almost 80%. Thus proving the memory of the event had not faded, but had not been summoned using the CB method. Petitmengin et al. (2013) suggest the results of Johansson et al. (2005) come from participants making use of their available information to justify their choice in the context of the present social interaction, which is based on a relationship of trust in the interviewer. Is it lasting? Would self-reflection, not just experimenter-enabled reflection, have made the results of Johansson et al. (2005) dissipate with time?

Such a concept would be similar to the Plott s (1996) discovered preferences . Plott (1996) wrote that through repetition and learning stable preferences would be discovered , having been unknown to the individual. Another intermediate position between stable and constructed preferences is offered by Hoeffler and Ariely (1999). They argue that individuals construct preferences after facing a new type of decision, but these preferences then stabilise with time. The Petitmengin et al. (2013) response to Johansson et al. (2005) could be interpreted favourably toward these positions.

Despite their evidence for bi-directionality between choice and preference and situational factors, Ariely and Norton (2008) did not advocate against a stable component of preferences. There is plenty of evidence supporting stability, even if imperfect. For example, aggregate choice profiles remain highly consistent in estimated discount rates over the two years of a field experiment by Meier and Sprenger (2014). Individuals correlations over time are also generally high, though some individuals show signs of instability. A longer study, lasting almost a decade, tested the stability of individuals risk and time preferences using experimental and survey measures (Chuang & Schechter, 2014). They found stable answers to survey questions, but experimental answers were less stable. How does stability coexist with the range of evidence calling for constructed preferences? The challenge appears to be in reconciling stable and constructed preferences, rather than in accepting one or the either.

The answer, might lie in understanding how individuals themselves see each situation. Our notion of stable preferences has thus-far reflected assumptions made by economists for mathematical considerations not psychological accuracy. Perhaps it is this economic concept of stable preferences which requires revision. Suppose individuals have a stable preference to do the right thing , while the preference is stable, observed behaviour would change according to context as the perception of the right thing changes. Asking participants on what basis they make decisions would enable us to construct a better picture of how they themselves see their preferences, and how they realise these preferences in each context. Such a thing is beyond the reach of traditional economics, and psychology. The latter maintains a taboo against the approach of introspection. Psychologists often cite articles, such as Johansson et al. (2005), to ridicule the notion of individuals accessing their own mental depths, as even the memory behind their choice seem to become misplaced after only a few moments. However, Petitmengin et al. (2013) demonstrate that memories of individuals decision-making process can be revisited. In asking participants afterwards to justify and explain their choice, Johansson et al. (2005) are gauging participants mechanic understanding of what led to their choice. An experiential understanding of our own minds is incompatible with this mechanistic understanding. It has been shown that the brain networks we use to think about our own experience, as well as that of others, have an antagonistic relationship with the brain networks we use to understand mechanism (Jack et al., 2012). Simply put, when we think of our mind as a mechanism it literally turns off the brain areas we use to understand our own experience, and vice versa. Petitmengin et al. (2013) is a serious attempt to re-access experience , whereas Johansson et al. (2005) is not. Of course, introspection does not offer a straightforward window to the mind , but to use Jack and Roepstorff s (2002) analogy: neither does fMRI. Understanding fMRI images requires consideration caution and background knowledge, but one would not dismiss its value. What is required, is the development of good methods. Petitmengin et al. s (2013) interview elicitation method, for instance, seems capable of re-accessing the experience leading up to choices, and is therefore a worthy method for investigating preferences. Before returning to this point, we shall look at Decision by Sampling (DbS), which offers a different understanding of preferences, and one that explains how preferences can be both stable and made up on the spot. We shall then evaluate whether a role remains for introspection.

DbS does not assume psychoeconomic functions exist, e.g. concave utility functions (Stewart, Chater & Brown, 2006). Instead, when confronted by a choice, individuals randomly compare the values of these options against a sample of attribute values. The distribution of attribute values from the problem, as well as distributions from recent and long-term memory, represent a source of samples. Similar to the completeness axiom, individuals are only assumed to tell if the target value is better or worse than the sampled value. Individuals keep a frequency count of these binary, ordinal comparisons, and express a preference once a threshold for favourable comparisons has been reached.

Ungemach, Stewart, and Reimers (2011) conducted a series of experiments where the shape of psychoeconomic functions, at least partly, seem to emerge from the distribution of attribute values in questions used to elicit them, as well as values held in short term memory. In Study 1a (2011) they offered customers leaving a supermarket the chance to swap their receipt for a chance to choose between two lotteries, one risky (but with higher pay-out) and the other safe (but with lower pay-out). They used the prices to calculate, for each individual, the difference in relative rank of the two lottery outcomes within the sample of the prices on their receipt. They found a positive association between the difference in relative rank and the probability of selecting the riskier lottery. In the same paper they also showed how thinking about one s upcoming birthday affected preferences in a choice between a smaller-sooner or larger-later amount. The sample values considered in these experiments did not offer an explanation of all decisions, but that is partly because samples from longer-term memories are beyond the access of the experimenters. It is possible, that some stable underlying mapping exists between objective and subjective values, but the effect of manipulations of attribute value distributions are at least as large as individual differences between people (Ungemach, Stewart, & Reimers, 2011). DbS argues residual preferences, might not exist, and that their appearance can be wholly explained by the stability of distributions of attribute values over time, held in long term memory (Stewart, Reimers & Harris, 2014). Then, does it not seem that we have stability in preferences, albeit in a different guise?

As with Ariely and Norton (2008), DbS does not seem able to rule out preferences as unstable . While under DbS comparisons are made on on the spot , the stability of attribute values in long term memory means beliefs and desires can be stable. Does DbS offer a good understanding of preferences? Or, is there anything to gain with introspection?

DbS has been tested in circumstances where choices had clear, quantitative dimensions. How would DbS fare in reality? Coca-Cola and Pepsi are nearly identical in composition, but the former is typically preferred over the latter. McClure et al. (2004) found that when the image of Coca-Cola preceded the drink, there was significantly more brain activity in the dorsolateral prefrontal cortex (DLPFC) and hippocampus. No similar activation existed for Pepsi, whether it was delivered with or without brand knowledge. The DLPRC and hippocampus have been associated with affective information processing and the recall of declarative memories. There are a few problems with this for DbS. What is being sampled? Qualitative dimensions, such as nostalgia , must exist in order to take account of the stronger preference for Coca-Cola than Pepsi, if there is no difference in taste. How are e.g. nostalgic impressions ranked? Secondly, DbS assumes dimensions are equally important. Suppose Coca-Cola beats Pepsi and Irn Bru on most dimensions, such price, taste, nostalgia, but Irn Bru, is Scottish , and just because of this dimension, Irn Bru is chosen consistently by this person. What if some dimensions are more important than other? DbS does not permit a discussion about the dimensions themselves. One favourable comparison is just that.

Besides the matter of quantifying qualitative dimensions, and the notion of equi-valuable dimensions, there is also, the matter of the comparison process. DbS is a descri ptive theory, which depends on pairwise, binary comparisons in the decision-making process. While areas of the brain associated to long-term memory and emotion were stimulated for Coca-Cola, Pepsi did not activate any brain area significantly more than without brand knowledge, as if they just were not making a comparison at all. One notable testing of DbS s predictions for the decision-making process has consisted of tests of eye-movement data such as Noguchi and Stewart (2014), where clear quantified values describe all the attributes of the options. What happens if the choice concern real-world options? Besides the McClure et al. study (2004), Petitmengin et al. s (2013) method of introspection also seemed to reveal a different decision-making process than envisioned by DbS. Beside switching between women s facial features, individuals also personified the photographs to make comparisons. For instance, they imagined the faces as belonging to a person, or they imagined them in a specific location, or they imagined an interaction with each woman. This comparison process is not like DbS because values are not being retrieved , impressions are being generated, and these are being compared based on feelings associated with them. Does the activation of the hippocampus for Coca-Cola (McClure et al., 2004) not suggest a similar process?

With Ariely and Norton (2008), we saw a picture of preferences that combined stable economic preferences with situational factors, such as context, environment and emotion, as well as choice-created preferences, explained with CD or SPT. Under Ariely and Norton s (2008) conception of preferences, beliefs and desires themselves are stable, but behaviour is also influenced by unstable components. With DbS, this notion of stable preference is replaced with a stable distribution of values drawn from long term memory. Preferences might be seen as stable, but compromised often, and by a variety of components as in Ariely and Norton (2008). Or, constructed moment-to-moment, but drawing on stable values from the past, as in DbS. We have suggested that a better understanding of preferences is required before the question can be answered, and that introspection might be useful toward this purpose. The interview elicitation method of Petitmengin et al. (2013) appears successful in its attempt to retrieve memories of the decision-making process. Combining this with fMRI could help in evaluating when introspection is being made up , e.g. Johansson et al. (2005), and when it is accurate, i.e. McClure et al. (2004).

References

Ariely, D., & Norton, M. I. (2008). How actions create not just reveal preferences. Trends in cognitive sciences, 12(1), 13-16.

Bem, D. J. (1972). Self-perception theory. Advances in experimental social psychology, 6, 1- 62.

Christakis, N. A., & Fowler, J. H. (2007). The spread of obesity in a large social network over 32 years. New England journal of medicine, 357(4), 370-379.

Christakis, N. A., & Fowler, J. H. (2008). The collective dynamics of smoking in a large social network. New England journal of medicine, 358(21), 2249-2258.

Chuang, Y., & Schechter, L. (2014). Stability of Experimental and Survey Measures of Risk, Time, and Social Preferences Over Multiple Years. Time, and Social Preferences over Multiple Years (March 18, 2014).

Crockett, M. J., Apergis-Schoute, A., Herrmann, B., Lieberman, M. D., M ller, U., Robbins, T. W., & Clark, L. (2013). Serotonin modulates striatal responses to fairness and retaliation in humans. The Journal of Neuroscience, 33(8), 3505-3513.

Dutton, D. G., & Aron, A. P. (1974). Some evidence for heightened sexual attraction under conditions of high anxiety. Journal of personality and social psychology, 30(4), 510.

Hall, L., Strandberg, T., P rnamets, P., Lind, A., T rning, B., & Johansson, P. (2013). How the polls can be both spot on and dead wrong: Using choice blindness to shift political attitudes and voter intentions. PloS one, 8(4), e60554.

Harmon-Jones, E. E., & Mills, J. E. (1999). Cognitive dissonance: Progress on a pivotal theory in social psychology. In Scientific Conferences Program, 1997, U Texas, Arlington, TX, US This volume is based on papers presented at a 2-day conference at the University of Texas at Arlington, winter 1997.. American Psychological Association.

Hoeffler, S., & Ariely, D. (1999). Constructing stable preferences: A look into dimensions of experience and their impact on preference stability. Journal of Consumer Psychology, 8(2), 113-139.

Jack, A. I., & Roepstorff, A. (2002). Introspection and cognitive brain mapping: from stimulus response to scri pt report. Trends in cognitive sciences, 6(8), 333-339.

Johansson, P., Hall, L., Sikstr m, S., & Olsson, A. (2005). Failure to detect mismatches between intention and outcome in a simple decision task. Science, 310(5745), 116- 119.

Johansson, P., Hall, L., T rning, B., Sikstr m, S., & Chater, N. (2014). Choice blindness and preference change: You will like this paper better if you (believe you) chose to read it!. Journal of Behavioral Decision Making, 27(3), 281-289.

Johnson, E. J., & Goldstein, D. G. (2003). Do defaults save lives? Science,302, 1338-1339.

McClure, S. M., Li, J., Tomlin, D., Cypert, K. S., Montague, L. M., & Montague, P. R. (2004). Neural correlates of behavioral preference for culturally familiar drinks. Neuron, 44(2), 379-387.

Meier, S., & Sprenger, C. D. (2014). Temporal stability of time preferences. Review of Economics and Statistics, [early access] doi:10.1162/REST_a_00433

Noguchi, T., & Stewart, N. (2014). In the attraction, compromise, and similarity effects, alternatives are repeatedly compared in pairs on single dimensions. Cognition, 132(1), 44-56.

Ormerod, P., & Wiltshire, G. (2009). Binge drinking in the UK: a social network phenomenon. Mind & Society, 8(2), 135-152.

Petitmengin, C., Remillieux, A., Cahour, B., & Carter-Thomas, S. (2013). A gap in Nisbett and Wilson s findings? A first-person access to our cognitive processes. Consciousness and cognition, 22(2), 654-669.

Plott, C. R. (1996). Rational individual behavior in markets and social choice processes: the discovered preference hypothesis (pp. 225-250). London, England: MacMillan.

Schwarz, N., & Clore, G. L. (1983). Mood, misattribution, and judgments of well-being: Informative and directive functions of affective states. Journal of personality and social psychology, 45(3), 513.

Stewart, N., Chater, N., & Brown, G. D. (2006). Decision by sampling. Cognitive psychology, 53(1), 1-26.

Stewart, N., Reimers, S., & Harris, A. J. (2014). On the origin of utility, weighting, and discounting functions: How they get their shapes and how to change their shapes. Management Science, 61(3), 687-705.

Tversky, A., & Simonson, I. (1993). Context-dependent preferences. Management science, 39(10), 1179-1189.

Ungemach, C., Stewart, N., & Reimers, S. (2011). How incidental values from our environment affect decisions about money, risk, and delay. Psychological Science, 22, 253-260.

Wansink, B. (2010). From mindless eating to mindlessly eating better. Physiology & Behavior, 100(5), 454-463.

Wansink, B., & van Ittersum, K. (2013). Portion size me: plate-size induced consumption norms and win-win solutions for reducing food intake and waste. Journal of Experimental Psychology: Applied, 19(4), 320.