ABSTRACT
We propose that the process of causal induction can be regarded as a form of belief-revision, and formalize this idea using a discrepancy-based learning algorithm similar to that employed in the Rescorla-Wagner model of associative learning (Rescorla and Wagner, 1972) and the Belief-Adjustment Model (Hogarth and Einhorn, 1992). We then demonstrate that this model can account for conflicting patterns in human induction judgments reported by Wasserman et al. (1993) and Buehner and Cheng (1997), two data-sets which it is difficult for other models to satisfactorily explain.
