# Towards Fairness in Personalized Ads Using Impression Variance Aware Reinforcement Learning

arXiv: https://arxiv.org/abs/2306.03293 · HTML: https://mlsanity.com/paper/2306.03293/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- Builds on 21 of its 41 references that are in the map.
- First posted on arXiv 5 Jun 2023.

## Builds on (most cited first, 21 of 21)

- [Equality of Opportunity in Supervised Learning](https://mlsanity.com/paper/1610.02413/) (2016)
- [A Survey on Bias and Fairness in Machine Learning](https://mlsanity.com/paper/1908.09635/) (2019)
- [On Fairness and Calibration](https://mlsanity.com/paper/1709.02012/) (2017)
- [Fairness Beyond Disparate Treatment & Disparate Impact: Learning Classification without Disparate Mistreatment](https://mlsanity.com/paper/1610.08452/) (2016)
- [A Practical Guide to Multi-Objective Reinforcement Learning and Planning](https://mlsanity.com/paper/2103.09568/) (2021)
- [Fairness without Demographics through Adversarially Reweighted Learning](https://mlsanity.com/paper/2006.13114/) (2020)
- [A comparative study of fairness-enhancing interventions in machine learning](https://mlsanity.com/paper/1802.04422/) (2018)
- [RUDDER: Return Decomposition for Delayed Rewards](https://mlsanity.com/paper/1806.07857/) (2018)
- [Fairness-Aware Ranking in Search & Recommendation Systems with Application to LinkedIn Talent Search](https://mlsanity.com/paper/1905.01989/) (2019)
- [Bias Mitigation for Machine Learning Classifiers: A Comprehensive Survey](https://mlsanity.com/paper/2207.07068/) (2022)
- [A Survey on the Fairness of Recommender Systems](https://mlsanity.com/paper/2206.03761/) (2022)
- [Robust Optimization for Fairness with Noisy Protected Groups](https://mlsanity.com/paper/2002.09343/) (2020)
- [Towards Long-term Fairness in Recommendation](https://mlsanity.com/paper/2101.03584/) (2021)
- [From Parity to Preference-based Notions of Fairness in Classification](https://mlsanity.com/paper/1707.00010/) (2017)
- [FairRec: Fairness-aware News Recommendation with Decomposed Adversarial Learning](https://mlsanity.com/paper/2006.16742/) (2020)
- [Reinforcement Learning with Trajectory Feedback](https://mlsanity.com/paper/2008.06036/) (2020)
- [Learning Long-Term Reward Redistribution via Randomized Return Decomposition](https://mlsanity.com/paper/2111.13485/) (2021)
- [Toward Pareto Efficient Fairness-Utility Trade-off inRecommendation through Reinforcement Learning](https://mlsanity.com/paper/2201.00140/) (2022)
- [Two-sided fairness in rankings via Lorenz dominance](https://mlsanity.com/paper/2110.15781/) (2021)
- [Discrimination through optimization: How Facebook's ad delivery can lead to skewed outcomes](https://mlsanity.com/paper/1904.02095/) (2019)
- [Towards Threshold Invariant Fair Classification](https://mlsanity.com/paper/2006.10667/) (2020)

## Built on by (most cited first, 6 of 6)

- [The Pursuit of Fairness in Artificial Intelligence Models: A Survey](https://mlsanity.com/paper/2403.17333/) (2024)
- [FairJob: A Real-World Dataset for Fairness in Online Systems](https://mlsanity.com/paper/2407.03059/) (2024)
- [Beyond Quality: Unlocking Diversity in Ad Headline Generation with Large Language Models](https://mlsanity.com/paper/2508.18739/) (2025)
- [Deep Reinforcement Learning for Ranking Utility Tuning in the Ad Recommender System at Pinterest](https://mlsanity.com/paper/2509.05292/) (2025)
- [Privacy-Preserving Race/Ethnicity Estimation for Algorithmic Bias Measurement in the U.S](https://mlsanity.com/paper/2409.04652/) (2024)
- [Productionized Fairness Measurement Under Privacy Constraints](https://mlsanity.com/paper/2606.27558/) (2026)

## Abstract (arXiv)

Variances in ad impression outcomes across demographic groups are increasingly considered to be potentially indicative of algorithmic bias in personalized ads systems. While there are many definitions of fairness that could be applicable in the context of personalized systems, we present a framework which we call the Variance Reduction System (VRS) for achieving more equitable outcomes in Meta's ads systems. VRS seeks to achieve a distribution of impressions with respect to selected protected class (PC) attributes that more closely aligns the demographics of an ad's eligible audience (a function of advertiser targeting criteria) with the audience who sees that ad, in a privacy-preserving manner. We first define metrics to quantify fairness gaps in terms of ad impression variances with respect to PC attributes including gender and estimated race. We then present the VRS for re-ranking ads in an impression variance-aware manner. We evaluate VRS via extensive simulations over different parameter choices and study the effect of the VRS on the chosen fairness metric. We finally present online A/B testing results from applying VRS to Meta's ads systems, concluding with a discussion of future work. We have deployed the VRS to all users in the US for housing ads, resulting in significant improvement in our fairness metric. VRS is the first large-scale deployed framework for pursuing fairness for multiple PC attributes in online advertising.