Stanford study: X's For You algorithm mistakes outrage for interest, drifting from what users value
StanfordHAI · x · 2026-08-19
A new Stanford-led study finds that X's "For You" algorithm mistakes outrage for interest—the better it gets at reading your behavior, the further your feed drifts from what you actually care about. The researchers documented a fundamental value misalignment between recommended content and users' stated personal values.
The root cause is engagement-based ranking: the algorithm heavily weights follows, time spent, reposts, and especially replies, using these signals as a proxy for interest. First author Ziv Epstein (postdoc in Michael Bernstein's lab) notes engagement is easy to measure, so platforms lean on it as the primary interest metric—but being rapt is not the same as wanting the content.
More from AGI Musings
- Building businesses that are both AI-Native and AI-Proof — julesyoo · 2026-08-19
- AI Hasn't Plateaued: Humans Pick Hypotheses, AI Does the Rigorous Work — Soulren · 2026-08-19
- Judea Pearl interview clips released — ArtificialOther · 2026-08-19
- Using LLMs daily makes your own reasoning feel weaker — two studies agree, causation unproven — Dangerous_Ad7101 · 2026-08-19
- Independent Minds Joining Frontier AI Labs: Reformers or Surrender? — chalmermagne · 2026-08-19
- AI slide decks all look the same, and polish is losing its signal value — AmbassadorSad3889 · 2026-08-19