A Bayesian Choice Model for Eliminating Feedback Loops

TL;DR

This paper introduces a Bayesian choice model that addresses feedback loops in personalization systems by accounting for limited user exposure, enabling fairer and more efficient online recommendation strategies.

Contribution

The paper presents a novel Bayesian choice model based on Luce axioms that explicitly models limited exposure and supports efficient, fair online recommendation with low regret.

Findings

Model accurately infers preferences from few interactions

Achieves low regret with minimal exploration

Supports fair presentation without bias against unpresented options

Abstract

Self-reinforcing feedback loops in personalization systems are typically caused by users choosing from a limited set of alternatives presented systematically based on previous choices. We propose a Bayesian choice model built on Luce axioms that explicitly accounts for users' limited exposure to alternatives. Our model is fair---it does not impose negative bias towards unpresented alternatives, and practical---preference estimates are accurately inferred upon observing a small number of interactions. It also allows efficient sampling, leading to a straightforward online presentation mechanism based on Thompson sampling. Our approach achieves low regret in learning to present upon exploration of only a small fraction of possible presentations. The proposed structure can be reused as a building block in interactive systems, e.g., recommender systems, free of feedback loops.