Skip to main content

Relative Rank Theory: A Computational Model of Preferences, Choices, Attitudes and Opinions

Periodic Reporting for period 1 - RRTJDM (Relative Rank Theory: A Computational Model of Preferences, Choices, Attitudes and Opinions)

Reporting period: 2018-10-01 to 2020-03-31

"Our differing preferences and attitudes make each of us unique. But how are our everyday choices related to our preferences, and how are the opinions that we express related to our underlying attitudes? In particular, how can we understand the phenomenon of opinion polarisation (as seen, for example, in people's opinions since the Brexit referendum in the UK), and how can we understand individual differences in people’s everyday choices when those choices are so strongly influenced by the context that the market provides?

The research is developing a radically new perspective on these and related questions, and is developing an integrative computational model which reconceptualises the relation between preferences, choices, attitudes, and expressed opinions.

Specifically, we assume that people’s choices and expressed opinions cannot be understood in terms of stable preferences and attitudes in the way that conventional economic models assume. Applying insights from social psychology to cognitive and economic models of choice, we are developing a novel alternative approach according to which underlying preferences and attitudes are stable characteristics of
people but cannot inform everyday behaviour directly because people have no conscious access to the strength of their underlying preferences and attitudes.

Extending the approach by applying cognitive modelling techniques to social network phenomena, we will quantify ""authenticity preference"" and ""social extremeness aversion"" and show how polarisation and contagion arise from the tension between these preferences.

The project has potential impact in a number of areas. Politicians seek policies that will satisfy people's preferences and maximise welfare. Allocation of resources (e.g. between new hospitals versus improved road safety)
often implicitly relies on ideas of maximizing people’s utility as inferred from their choices. Our approach undermines the existing ways of approaching these issues, and also offers a new perspective on important
contemporary social phenomena such as polarisation and social contagion effects."
During the first reporting period our main activity has been the development of two theoretical models. Each model embodies two of our key assumptions and shows how each assumption enables understanding of phenomena that are otherwise difficult to explain. The first assumption is that there is an important psychological distinction between our underlying preferences and attitudes and our beliefs about what those preferences and attitudes are. The second assumption is that our preferences and attitudes must be represented as distributions rather than single points.

The first model is a cognitive model of social influence (Social Sampling Theory: SST). The model is applied to several aspects of social network behaviour, with a particular emphasis on political polarization and contagion effects in social networks. People in a social network are assumed to observe the behaviour of their network neighbours and thereby infer the social distribution of particular attitudes and behaviours. It is assumed in SST that (a) people dislike behaving in ways that are extreme within their neighbourhood social norm (social extremeness aversion assumption), and hence tend to conform and (b) people prefer to behave consistently with their own underlying attitudes (authenticity preference assumption) hence minimizing dissonance. People’s preferences for authenticity (effectively, being true to themselves) will often be conflict with their desire to fit in with their social group, and hence the attitudes that people choose to express well, according to the model, represent a compromise between these two conflicting tendencies.

We show that SST provides a new perspective on a number of well-known phenomena including for example (a) the development of segregated opinion neighborhoods and echo chambers, (b) political polarization, and (c) the opposing effects on subjective well-being of authentic behavior and high levels of social comparison.

The second model (Relative Rank Theory: RRT) aims to reconcile the systematic context-dependence of choice with the existence of stable individual differences in people’s preferences. RRT distinguishes between two types of psychological preference. Underlying preferences (preferencesU) are stable across contexts but do not directly inform everyday choices from markets of three or more options; expressed preferences (preferencesE, which can be thought of as beliefs about preferences) are context-dependent, learned, and do directly inform such choices. Together with the assumption that judgements and valuations are relative and rank-based, this distinction enables RRT to account for stability and context-dependence in choice without assuming commensurability of attribute values. The basic architecture of RRT is shown in the attached figure in the context of preferences for, and choices between, cheeses varying in sharpness.

We have also examined the more general assumptions that underlie our approach. One central assumption is that people are sensitive to the relative ranked positions that they occupy within a social comparison group. For example, people’s subjective well-being is determined by how their income ranks within a social comparison group rather than with their income per se. In additional work using existing datasets we have been able to confirm a prediction of this approach, showing in two large international samples that a 10% increase in income is associated with an increase in life satisfaction that is more than twice as large in a country with low-income inequality as it is in a country with high income inequality (Quispe-Torreblanca et al., in press). We have also explored the theoretical relationship between our rank-based models of subjective well-being and various indices of relative deprivation (Hounkpatin, Wood, & Brown, 2020) and have also found (consistent with the idea that people care about their perceived social rank) that increases in life satisfaction are associated with increases in conspicuous consumption rather than with increases in income (Brown & Gathergood, 2020).

Brown, G. D. A., & Gathergood, J. (2020). Consumption changes, not income changes, predict changes in subjective well-being. Social Psychological and Personality Science, 11, 64-73.

Hounkpatin, H. O., Wood, A. M., & Brown, G. D. A. (2020). Comparing indices of relative deprivation using behavioural evidence. Social Science & Medicine.

Quispe-Torreblanca, E. G., Brown, G. D. A., Boyce, C. J., Wood, A. M., & De Neve, J. (in press). Inequality and social rank: Income increases buy more life satisfaction in more equal countries. Personality and Social Psychology Bulletin.
The theoretical models that we have developed have implications for societal issues such as political polarisation and market efficiency. For example, our social sampling model helps us understand when polarisation is more or less likely to occur; our model of inferred preferences sets limits on the extent to which free trading (as in liberalised markets) will be efficient in the sense of satisfying people’s underlying preferences.

Much of the research in the remainder of the research period will be devoted to exploring and illustrating these implications as well as to testing more specific predictions of the model.