What are Choice Prediction Competitions?

Choice prediction competitions aim to promote the development of models that may predict, rather than merely describe, behavior in the social sciences. Each competition focuses of a different area of interest in behavioral decision/economic research.

each competition starts with a large experiment which includes many choice problems, randomly selected from the space of relevant problems. The results of this experiment are then posted on the competition website, and we invite researchers to fit models on this data. We then challenge them to submit their models to compete on the prediction of the results of another large experiment (the competition set) that includes similar but different randomly selected problems from the relevant space. Thus they focus on predicting behavior, rather than merely fitting previous results.

The projects were initiated by Ido Erev, Al Roth, and myself, and have been funded by the BSF. This page summarizes the competitions we ran thus far, for more data see our competitions websites and papers. In 2010 the new journal GAMES published a special issue on“predicting behavior in games” that includes the 2nd and 3rd competition described in this page, as well as papers by colleagues who participated in these competitions.