Proceedings of the Annual Meeting of the Cognitive Science Society

Reference resolution is one of the core components of language understanding. In spite of its centrality, psychological evidence has shown that the reference resolution process is prone to errors and egocentric bias. In this work, we propose an extension to Analogical Reference Resolution, a computational model based on analogical retrieval, which accounts for such errors. We test the extended model on a study by Epley et al. (2004) and replicate human patterns of bias and correction.