Discovering Quantification and Number in a Role-Filler Model

Quantification plays a central role in human reasoning and models thereof, but the discovery and development of quantification remains an open question. We present a theory of how such concepts are learned from experience in the DORA model, a neurally-plausible computational model of relational learning and reasoning (Doumas et al., 2008). The same theory accounts for how concepts of number are acquired in this class of model. We are unaware of any prior model that accounts for the development of both quantification and number from unstructured (e.g.,