The Rationality of Prejudices

People frequently judge and discriminate each other on the basis of their skin color, gender, or clothing style. Beyond the ethical issues they raise, such prejudices seem inefficient. They often lead to erroneous judgements, missed opportunities and resentment and, at the aggregate level, to segregation, riots, or religious conflicts. Why, then, are they so ubiquitous?

In this paper, we argue that prejudices strive because strategies that rely on them can be very successful in competitive environments. Prejudices are heuristics based on accumulated experiences, expressed as simple cognitive relationships between specific traits (e.g., height) and behavior (e.g., aggressiveness). These associations, acquired through evolution or experience, enable people to reach judgments about complex situations or competitors in a “blink”. In this sense, they are closely related to people’s ability to process large amounts of information rapidly and often unconsciously, and to reach quick decisions.

One explanation for these “heuristics that make us smart” is evolutionary. When our ancestors interacted with strangers, those who could rapidly and accurately discriminate between dangerous and trustworthy partners were more likely to survive and reproduce. However, this explanation does not inform us about the conditions under which relying on these cognitive shortcuts is rational or optimal. The fact that, empirically, people do rely on rules of thumb certainly implies that these rules can be useful and efficient, but not that they always are. In fact, even though our intuitive judgments are often accurate, they also frequently lead to errors and inaccuracies, so that the opposite argument could be made equally well: evolutionarily, those most likely to survive are those able to assess situations calmly and to derive rational conclusions–not hasty responses based on emotions or “gut-feelings”.

In summary, we know little about the fitness of strategies that use only a limited subset of the available information to reach conclusions about their social partners. Can prejudices–the extrapolation onto others of the behavior of people characterized by similar attributes–form the basis of a successful strategy in a competitive environment? In this paper, we investigate the performance of these rules of thumb by putting them in competition with well-known strategies such as Tit-For-Tat or defection. We derive conditions under which prejudiced strategies outperform these other strategies and when, on the contrary, they are suboptimal.

Prejudices have the advantage of providing pre-defined guidelines for interactions, without the need to learn the other’s specificities. As such, they enable rapid reactions to unknown circumstances–for example those involving a significant portion of interactions with foreigners. We will show that they can successfully avoid exploitation, while still taking advantage of cooperation with populations that have been found to be cooperative. Of course, such learning speed comes at a cost. Because prejudices are coarse-grained inferences based on a limited number of attributes, they are particularly prone to error (e.g., not all green people are uneducated) and hence lead to some level of exploitation (e.g., wrongly assuming that all blue people are cooperators) or missed opportunities.

Despite this inaccuracy, we find that strategies based on prejudices perform well for a large range of parameters. They are particularly well suited for situations in which the population renews itself relatively rapidly–for example because of high migration or birth rates. In these situations, they even outperform the most successful strategies that have previously been proposed in the literature [15]: those based on reciprocity (Tit-For-Tat, a strategy in which a player starts cooperating, and then copies the interaction partner’s decisions), and those based on exploitation (ALLD, in which an individual always defects).

By: Thomas Chadefaux, Dirk Helbing

Chair of Sociology, in particular of Modeling and Simulation, Swiss Federal Institute of Technology Zurich, Zurich, Switzerland

Category(s): Uncategorized

Leave a Reply