Beliefs about people's prosociality: Eliciting predictions in dictator games

One of the most pervasive economic decisions that people have to take is whether to enter an economic interaction. A rational decision process takes into account the probability that the partner will act in a favorable way, making the interaction or the cooperative activity beneficial. Do people actually decide upon such predictions? If yes, are these predictions accurate? We describe a novel experimental method for eliciting participants' implicit beliefs about their partners' prosociality: In a modified dictator game, receivers are offered to forgo what the dictator shall transfer and take a sure amount instead. We then infer receivers' subjective probabilities that the dictator makes a prosocial decision. Our results show that people do form prior beliefs about others' actions based on others' incentives, and that they decide whether to enter an interaction based on these beliefs. People know that others have prosocial as well as selfish preferences, yet the prior beliefs about others' prosocial choices is biased: First, participants underestimate others' prosociality. Second, their predictions about others' choice correlate with their own choice, reflecting a consensus effect. We also find a systematic difference between implicit and explicit predictions of others' choices: Implicit beliefs reflect more trust towards others than explicit statements.