Evaluating Instance-based Learning in Multi-cue Diagnosis


Decision heuristics are often described as fast and frugal, meaning that they take little time and require relatively few computations to make a decision when compared to optimal decision systems (Gigerenzer & Todd, 1999). Fast & Frugal Trees are one heuristic that are a special case of decision trees in which there is a possible exit out of the decision process at every cue considered in the tree (Luan, Schooler, & Gigeren- zer, 2011). There is currently no computational account of how hu- mans learn heuristics like F&FT-based decision processes. This is a significant gap in our scientific understanding, and we aim to begin addressing that gap in this effort. In this ab- stract we report results from a pilot study assessing Instance- based Learning Theory (IBLT) as an account of human learn- ing from experience in domains where F&FTs may be good decision heuristics, such as diagnostic tasks.