Simulating Feature- and Relation-Based Categorisation with a Symbolic-Connectionist Model

AbstractParticipants in Goldwater et al. (2018) reported using either feature- or relation-based strategy during a series of category learning tasks. A computational modeling study was conducted to investigate whether performance on Experiments 1 and 2 of Goldwater et al. (2018) might be explained by the assumption that participants used either feature- or relation-based representational encoding during learning. Human participants' and model performance are compared and implications are discussed.

Return to previous page