Hagmayer, Y.
University of Goettingen
Most theories of repeated decision making do not take causal considerations into account. The two most prominent theories assume that people either learn about exemplars or abstract linear rules from the observations made (e.g., Juslin et al., 2003). Causal model theories, by contrast, claim that people acquire and use knowledge about the causal structure underlying a decision problem to decide on interventions (Sloman & Hagmayer, 2006). An experiment was conducted combining paradigms on multiple cue and causal learning. Participants were confronted with a fictitious medical system consisting of four factors contributing to an outcome. Participants’ assumptions about the causal relations between the four factors were manipulated by instruction in three groups. In a learning phase participants received identical learning input about the state of the four factors and the value of the outcome before and after an intervention on one of the factors. The learning input consisted of 110 trials and was equally compatible with all causal assumptions. In two test phases (before and after the learning phase) participants were requested to estimate the value of the outcome based on the observed factors, choose one of two interventions, and predict the value of the resulting outcome. Learning input and test cases were constructed to allow differentiating the three theoretical accounts. The results showed that participants considered causal structure even after extensive learning experience. Depending on their causal assumptions, participants preferred different interventions. To further analyze the data, an exemplar model, a linear model and a causal-model model were fitted to participants’ answers. The model assuming that participants induce a causal model and update its parameters based on the learning input predicted participants’ judgments and choices best. These findings indicate that people may also use causal reasoning as a decision making strategy.