How do people assign responsibility to an individual cause or person in a situation of collective responsibility? This dissertation addresses the question by designing an experimental game, the Triangle Game (TG), which participants play in a group with three computer players. The participants’ task is to count triangles presented in complex diagrams for a brief period of time. Whether the group wins or loses depends on the accuracy of each player in the group. After each round, participants assign responsibility for the result to each player. Three experimental conditions differ in how the individual judgments are combined to determine the deviation of the group’s answer from the correct solution. This deviation determines whether a round is lost or won. For the three experimental conditions, the group’s deviation is determined, respectively, by the sum of each player’s deviation, the deviation of the least accurate player or the deviation of the most accurate player. The results show that these differences in the underlying causal structure have a significant influence on participants’ responsibility ratings. Furthermore, the predictions of different cognitive models are tested. A model for assigning responsibility to individual causes in cases of multiple causation developed by Chockler and Halpern (2003) describes the empirical data best. A second experiment replicates these findings. In an additional step, participants were asked to change the result of each round, for example from a loss to a win, by minimally altering the answers that the players had given in that round. The results show that participants perceive a possible world, where several small changes have been made as more similar to the actual world than a possible world with one big change. The implications of these findings for counterfactual theories of causation are discussed. The main advantage of the TG is that it allows testing psychological theories of responsibility attribution in a formally rigorous manner.
<< Back to list of publications