Proceedings of the Annual Meeting of the Cognitive Science Society

We investigate a method for formulating context- and task- specific computational models of human performance in a con- strained semantic memory task. In particular, we assume that memory retrieval can only use a simple process – a random walk – and examine whether the effect of context and task specifications can be captured via a straightforward network estimation method that is sensitive to context and task. We find that a random walk model on the context-specific networks mimics aggregate human performance.