Output details

Evaluating the inverse decision-making approach to preference learning

<22> Originality: First paper to systematically evaluate a generative or "inverse decision-making" model of preference learning.

Significance: Shows that generative models of preference learning predict detailed patterns in human inferences. Reveals that human learners infer preferences using several sources of information from choice events (e.g., diversity and valence of features of chosen and foregone options), and combine that information in a way that is difficult to capture using discriminative models.

Rigour: Includes a detailed analysis of generative and alternative models across multiple experiments, considering accuracy and parsimony. Published in a peer-reviewed conference with a low acceptance rate (25%).