Interactive preference elicitation under noisy preference models: An efficient non-Bayesian approach

Guillaume Escamocher, Samira Pourkhajouei, Federico Toffano, Paolo Viappiani, Nic Wilson

Abstract: Highlights•We develop a non-Bayesian approach for incremental elicitation of preferences and recommendation with a noisy user model.•Our approach is based on reasoning about the most plausible user preference models.•Results show, including in real-time contexts, the high accuracy of our method in returning the user's favoured alternative.•We prove formal asymptotic results for our algorithm, including the probability that it returns the actual best option.