Proceedings of the Annual Meeting of the Cognitive Science Society

The paper explores how people revise their belief in ahypothesis and the reliability of sources given independenceof sources or partial dependence (e.g. the sources share abackground). Specifically, we test a formal model ofreliability revision.The study provides support for Bovens and Hartmann’smodel of reliability revision. If a source provides a positivereport for an unlikely hypothesis, participants initially revisethe reliability of the source negatively. However, as additionalpositive reports emerge, they increase their estimate of thereliability of the source. Further, if it becomes known thesources are partially dependent (here, taught the same schoolof thought), the reliability of the source decreases again. Bothof these findings are in line with the Bayesian approach toreliability revision.