Rawlsian Explorer Recommenders: Minority Discovery Budgets for LLM-Based Personalization

LLM-based recommenders can entrench mainstream content (Sakib & Das, 2024) and polarize news exposure (Tsekhmeistruk, 2024). We propose a contextual bandit framework with a Rawlsian minimax objective (Barsotti & Koçer, 2022): set an explicit “discovery budget” that prioritizes worst-off content segments (e.g., niche genres, minority creators), ensuring sufficient exploration to learn their true quality. Unlike ad hoc diversity heuristics, this yields a welfare-grounded optimization that seeks to improve the least advantaged slice of the catalog. We incorporate acceptance-aware presentation, informed by Juijn et al. (2023), surfacing process explanations and prioritizing reductions in “false negatives” for niche content (i.e., missed opportunities)—a preference users empirically show. We also compare a bias-correcting mode (actively counterbalancing historical skew) with a bias-preserving mode (Urchs et al., 2025) when representational fidelity is a desideratum, assessing trade-offs transparently. Evaluation spans content diversity, subgroup exposure, and satisfaction; we also audit how SES context modulates bias (Sakib & Das, 2024). Challenging the presumed utility–fairness trade-off (De-Arteaga et al., 2022), we test whether well-calibrated exploration improves both long-term utility (via better estimates of minority content quality) and equity. If successful, this offers a principled, legally defensible path for platforms to improve representational equity without sacrificing user experience.

References:

1. Challenging Fairness: A Comprehensive Exploration of Bias in LLM-Based Recommendations. Shahnewaz Karim Sakib, Anindya Bijoy Das (2024). BigData Congress [Services Society].
2. QUANTIFYING ALGORITHMIC BIAS IN NEWS RECOMMENDATIONS: METHODOLOGIES AND CASE STUDIES. Roman Tsekhmeistruk (2024). Scientific Journal of Polonia University.
3. MinMax fairness: from Rawlsian Theory of Justice to solution for algorithmic bias. F. Barsotti, R. G. Koçer (2022). Ai & Society.
4. Perceived Algorithmic Fairness using Organizational Justice Theory: An Empirical Case Study on Algorithmic Hiring. Guusje Juijn, Niya Stoimenova, João Reis, Dong Nguyen (2023). AAAI/ACM Conference on AI, Ethics, and Society.
5. Are All Genders Equal in the Eyes of Algorithms? - Analysing Search and Retrieval Algorithms for Algorithmic Gender Fairness. Stefanie Urchs, Veronika Thurner, M. Aßenmacher, Ludwig Bothmann, Christian Heumann, Stephanie Thiemichen (2025). arXiv.org.
6. Algorithmic fairness in business analytics: Directions for research and practice. Maria De-Arteaga, S. Feuerriegel, M. Saar-Tsechansky (2022). Production and operations management.