No abstract

No abstract

Recent experimental research on ultimatum and dictator games has found that first movers in such games tend to offer more to their counterparts than noncooperative game theory would predict. In fact, the modal offer is generally half the pie to be divided, while noncooperative game theory would suggest an offer of the smallest monetary unit. It is often argued that these results suggest a taste for fairness on the part of students participating in these experiments. In this paper we report the results of ultimatum and dictator games experiments designed to explore the underlying reasons for this apparent taste for "fairness." We find that if the right to be the first mover is earned by scoring high on a general knowledge quiz, and that right is reinforced by the instructions as being earned, then first movers behave in a significantly more self-regarding manner. Because our instructional procedures for earning rights can be interpreted as a "demand" treatment, but also to remove ail social influences on choice, we conducted "double blind" dictator experiments, in which individual subject decisions cannot be known by the experimenter or by anyone except the decision maker. The results yielded by far our largest observed number of selfregarding offers --significantly more than obtained in any of our other treatments: or any previously reported in the literature. Our interpretation is that offers in ultimatum and dictator games appear to be determined predominantly by strategic and expectations considerations. Other-regarding behavior is primarily an expectations phenomenon, rather than the result of an autonomous private preference for equity.

Cooperation between individuals requires the ability to infer each other's mental states to form shared expectations over mutual gains and make cooperative choices that realize these gains. From evidence that the ability for mental state attribution involves the use of prefrontal cortex, we hypothesize that this area is involved in integrating theory-of-mind processing with cooperative actions. We report data from a functional MRI experiment designed to test this hypothesis. Subjects in a scanner played standard two-person ''trust and reciprocity'' games with both human and computer counterparts for cash rewards. Behavioral data shows that seven subjects consistently attempted cooperation with their human counterpart. Within this group prefrontal regions are more active when subjects are playing a human than when they are playing a computer following a fixed (and known) probabilistic strategy. Within the group of five noncooperators, there are no significant differences in prefrontal activation between computer and human conditions.