Alexandria: A Library of Pluralistic Values for Realtime Re-Ranking of Social Media Feeds

2025-05-16

AI Summary

Current social media ranking algorithms over-rely on engagement metrics, neglecting multidimensional values such as user well-being and information diversity, thereby failing to accommodate heterogeneous user needs. Method: We propose PluralRank—a framework featuring a plug-and-play value library encompassing 78 interdisciplinary values; an LLM-driven content classifier integrated with a browser extension architecture for client-side, real-time streaming re-ranking; and a novel human-in-the-loop value configuration interface enabling fine-grained, personalized preference specification. Contribution/Results: PluralRank breaks from the dominant single-metric optimization paradigm by systematically engineering pluralistic values into composable, deployable ranking modules. A user study with 269 participants demonstrates statistically significant improvements in value alignment (+32.7%) and perceived control (+41.5%), validating the feasibility and effectiveness of pluralistic value integration in real-world social platforms.

Technology Category

• Machine Learning: Learning Preferences or Rankings
• Humans and AI: Learning Human Values and Preferences
• Application Domains: Humanities & Computational Social Science

Application Category

• User Modeling, Personalization and Recommendation: Personalized content ranking and presentation
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Social media feed ranking algorithms fail when they too narrowly focus on engagement as their objective. The literature has asserted a wide variety of values that these algorithms should account for as well -- ranging from well-being to productive discourse -- far more than can be encapsulated by a single topic or theory. In response, we present a $ extit{library of values}$ for social media algorithms: a pluralistic set of 78 values as articulated across the literature, implemented into LLM-powered content classifiers that can be installed individually or in combination for real-time re-ranking of social media feeds. We investigate this approach by developing a browser extension, $ extit{Alexandria}$, that re-ranks the X/Twitter feed in real time based on the user's desired values. Through two user studies, both qualitative (N=12) and quantitative (N=257), we found that diverse user needs require a large library of values, enabling more nuanced preferences and greater user control. With this work, we argue that the values criticized as missing from social media ranking algorithms can be operationalized and deployed today through end-user tools.

Problem

• Social media algorithms narrowly focus on engagement, ignoring diverse values
• Existing literature suggests 78 pluralistic values for better feed ranking
• End-user tools can operationalize missing values in real-time re-ranking

Innovation

• Library of 78 pluralistic values for algorithms
• LLM-powered classifiers for real-time feed re-ranking
• Browser extension enabling user-selected value prioritization