Towards Fair Personalization by Avoiding Feedback Loops
This work addresses the problem of unfair personalization for users of recommender systems, where certain content is over or under-presented.
This paper investigates how self-reinforcing feedback loops in recommender systems lead to biased user preference estimates, specifically overestimating over-presented content and underestimating censored alternatives. Through simulations, the authors demonstrate that simply conditioning on limited exposure can mitigate these biases.
Self-reinforcing feedback loops are both cause and effect of over and/or under-presentation of some content in interactive recommender systems. This leads to erroneous user preference estimates, namely, overestimation of over-presented content while violating the right to be presented of each alternative, contrary of which we define as a fair system. We consider two models that explicitly incorporate, or ignore the systematic and limited exposure to alternatives. By simulations, we demonstrate that ignoring the systematic presentations overestimates promoted options and underestimates censored alternatives. Simply conditioning on the limited exposure is a remedy for these biases.