Social prejudice has been defined as a cognitive error (false generalisation) directed towards a whole group or one of its members simply because of his or her affiliation. Emphasis has often been laid on the static nature of prejudice, refractory to falsification. However, in a number of experiments (cf. [4]), prejudice is found to change over time. Furthermore, prejudice is not necessarily negative: it can also be positive, although with a lower frequency. Why, what are the effects and functions of prejudice? What are the connections between prejudice, on one hand, and more general social evaluation, on the other? Here is where agent-based computer simulation of social phenomena can help. For both ethical and technical reasons, it is extremely difficult to carry on experiments with natural subjects on the global effects of prejudice. It is necessary to transfer the cognitive model into an agent-based simulation model, to implement it on a multi-agent platform, and to recursively refine the model on the grounds of the experimental findings. What we present here is a set of results from a simulation experiment based on the above questions, in the hope that it can provide a basis for bettering our understanding of and our insight on prejudice.
