Bonsai: Intentional and Personalized Social Media Feeds

2025-09-12

AI Summary

Social media recommendation systems over-optimize for engagement, causing content distribution to diverge from users’ authentic informational intentions. To address this, we propose Bonsai—the first end-to-end explainable feed-generation system supporting natural-language intention specification. Bonsai explicitly models user intent as the primary driver of feed construction, decoupling engagement objectives from underlying content preferences. It enables platform-agnostic, modular control—spanning planning, acquisition, curation, and ranking—while ensuring transparency throughout the pipeline. Natural-language interfaces and stepwise process visualization jointly enhance interpretability and user controllability. In a user study with 15 Bluesky participants, Bonsai significantly improved novel content discovery and harmful content filtering. Crucially, its transparency directly increased user trust and self-reported continued usage intent.

Technology Category

• Natural Language Processing: Generation
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Machine Learning: Learning Preferences or Rankings

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Web Mining and Content Analysis: Content-based information diffusion

Abstract

Modern social media feeds use predictive models to maximize engagement, often misaligning how people consume content with how they wish to. We introduce Bonsai, a system that enables people to build personalized and intentional feeds. Bonsai implements a platform-agnostic framework comprising Planning, Sourcing, Curating, and Ranking modules. Altogether, this framework allows users to express their intent in natural language and exert fine-grained control over a procedurally transparent feed creation process. We evaluated the system with 15 Bluesky users in a two-phase, multi-week study. We find that participants successfully used our system to discover new content, filter out irrelevant or toxic posts, and disentangle engagement from intent, but curating intentional feeds required participants to exert more effort than they are used to. Simultaneously, users sought system transparency mechanisms to trust and effectively use intentional, personalized feeds. Overall, our work highlights intentional feedbuilding as a viable path beyond engagement-based optimization.

Problem

• Addresses misalignment between engagement-driven feeds and user intent
• Enables personalized social media feeds through natural language expression
• Provides procedural transparency and user control in feed creation

Innovation

• Platform-agnostic framework with modular design
• Natural language intent expression for feed customization
• Procedurally transparent feed creation process control