Table of Contents
Fetching ...

The Relevance of Item-Co-Exposure For Exposure Bias Mitigation

Thorsten Krause, Alina Deriyeva, Jan Heinrich Beinke, Gerrit York Bartels, Oliver Thomas

TL;DR

The paper tackles exposure bias in implicit-feedback recommender systems, focusing on how item co-exposure and choice-set composition influence logged interactions. It evaluates MNL against other discrete-choice models under controlled human data, comparing exposure from overexposure and competition and assessing the necessity of logging observed alternatives. Key findings show that discrete-choice models mitigate exposure bias by conditioning on choice sets and observed alternatives, with multivariate variants uniquely robust to competition; differences among discrete-choice specifications are not clearly significant. The work suggests practical implications for logging policies and model training, highlighting substantial potential impact on recommendation quality while acknowledging the need for replication on real-world data and consideration of computational costs.

Abstract

Through exposing items to users, implicit feedback recommender systems influence the logged interactions, and, ultimately, their own recommendations. This effect is called exposure bias and it can lead to issues such as filter bubbles and echo chambers. Previous research employed the multinomial logit model (MNL) with exposure information to reduce exposure bias on synthetic data. This extended abstract summarizes our previous study in which we investigated whether (i) these findings hold for human-generated choices, (ii) other discrete choice models mitigate bias better, and (iii) an item's estimated relevance can depend on the relevances of the other items that were presented with it. We collected a data set of biased and unbiased choices in a controlled online user study and measured the effects of overexposure and competition. We found that (i) the discrete choice models effectively mitigated exposure bias on human-generated choice data, (ii) there were no significant differences in robustness among the different discrete choice models, and (iii) only multivariate discrete choice models were robust to competition between items. We conclude that discrete choice models mitigate exposure bias effectively because they consider item-co-exposure. Moreover, exposing items alongside more or less popular items can bias future recommendations significantly and item exposure must be tracked for overcoming exposure bias. We consider our work vital for understanding what exposure bias it, how it forms, and how it can be mitigated.

The Relevance of Item-Co-Exposure For Exposure Bias Mitigation

TL;DR

The paper tackles exposure bias in implicit-feedback recommender systems, focusing on how item co-exposure and choice-set composition influence logged interactions. It evaluates MNL against other discrete-choice models under controlled human data, comparing exposure from overexposure and competition and assessing the necessity of logging observed alternatives. Key findings show that discrete-choice models mitigate exposure bias by conditioning on choice sets and observed alternatives, with multivariate variants uniquely robust to competition; differences among discrete-choice specifications are not clearly significant. The work suggests practical implications for logging policies and model training, highlighting substantial potential impact on recommendation quality while acknowledging the need for replication on real-world data and consideration of computational costs.

Abstract

Through exposing items to users, implicit feedback recommender systems influence the logged interactions, and, ultimately, their own recommendations. This effect is called exposure bias and it can lead to issues such as filter bubbles and echo chambers. Previous research employed the multinomial logit model (MNL) with exposure information to reduce exposure bias on synthetic data. This extended abstract summarizes our previous study in which we investigated whether (i) these findings hold for human-generated choices, (ii) other discrete choice models mitigate bias better, and (iii) an item's estimated relevance can depend on the relevances of the other items that were presented with it. We collected a data set of biased and unbiased choices in a controlled online user study and measured the effects of overexposure and competition. We found that (i) the discrete choice models effectively mitigated exposure bias on human-generated choice data, (ii) there were no significant differences in robustness among the different discrete choice models, and (iii) only multivariate discrete choice models were robust to competition between items. We conclude that discrete choice models mitigate exposure bias effectively because they consider item-co-exposure. Moreover, exposing items alongside more or less popular items can bias future recommendations significantly and item exposure must be tracked for overcoming exposure bias. We consider our work vital for understanding what exposure bias it, how it forms, and how it can be mitigated.
Paper Structure (9 sections, 5 figures)

This paper contains 9 sections, 5 figures.

Figures (5)

  • Figure 1: Dataset structure.
  • Figure 2: Bias from overexposure from krause2024mitigating.
  • Figure 3: Bias from competition from krause2024mitigating.
  • Figure 4: Measured nDCG when training on data with uniform exposure frequencies and overexposure from krause2024mitigating.
  • Figure 5: Measured nDCG when training on data with popular and unpopular competition from krause2024mitigating.