Tournesol: YouTube Preference Learning with Bradley-Terry Models (2026)

Tournesol provides preference-learning and social-choice data structures for YouTube recommendations, developed by the tournesol-app organization and updated through March 2026. The platform implements Bradley-Terry models and preference aggregation to rank content based on AI ethics. It serves as an open-source framework for community-driven video evaluation.

Use Cases

• Training Bradley-Terry models to rank content based on pairwise preference-learning labels
• Implementing golden-ratio-optimization for preference-aggregation in social choice systems
• Building recommendation-engine logic that prioritizes AI ethics for YouTube video selection

Strengths

• Implements Bradley-Terry models for preference learning
• Includes golden-ratio-optimization for preference aggregation
• Provides Django REST framework integration for recommendation-engine deployment

Limitations

• Total row count and specific dataset dimensions are not provided in the source metadata
• Requires familiarity with Python and Django to navigate the platform's data structures

Provenance

Source

tournesol-app via GitHub

Collection Method

Preference-learning via user interaction on the Tournesol platform

Freshness

Updated through March 2026