“Our feeds shouldn't value confrontational engagement over complementary content, nor should they obsequiously feed us only content we agree with. They should strike a balance and mend social disagreements rather than make them worse.”
— Michael Bernstein, Stanford HAI Senior Fellow
— Michael Bernstein, Stanford HAI Senior Fellow
Recommender Objective Matrix
Interactive Scoring Weights
Active User Profile
Ideological Lean: -0.65 (Left/Progressive AI Policy) | Tolerance: 0.40
Discourse Telemetry
Polarization Δ
-0.28
Negative = Reduced Affective Polarization
Cross-Cutting Exp.
0.42
Outgroup viewpoint exposure ratio
Consensus Retention
0.84
Projected audience dwell stability
Effective Diversity
0.79
Entropy across distinct viewpoints
Feed Ideological Distribution
Methodological Grounding & Limitations
This simulator operationalizes Michael Bernstein's bridging formulation by rewarding posts endorsed across ideological splits (high cross-cutting consensus) rather than solely optimizing for click/reply dwell times that correlate with outrage. Limitation: Bridging metrics require sufficient signal on outgroup endorsement, which can suffer cold-start constraints in smaller micro-communities.