Top-K Contextual Bandits with Equity of Exposure

Abstract

Probability Ranking Principle 에 의하면 top-K items 을 greedy 하게 rank 하는것이 optimal 함

대신

Introduction
- This work investigates how the “equity of exposure” principle can be applied to top-K contextual bandit problems for recommendation
- We propose a novel algorithm for Exposure-Aware aRm Selection(EARS), which tackles the relevance-fairness trade-o” in a personalised manner.
  - To the best of our knowledge, EARS is the first algorithm that deals with equity of exposure in top-K contextual bandit recommenders.
Related paper
- Evaluating Stochastic Rankings with Expected Exposure: equity of exposure principle 제안, exploiting the stochasticity of rankings to lead to fairness in expectation.
- Investigating Listeners’ Responses to Divergent Recommendations: how different users respond to diverging recommendations, and observed that the openness of a user to randomised recommendations can vary wildly