RecExplain

RecExplain is a web app (with optional browser extension) that helps product teams and compliance stakeholders understand and steer their recommendation systems. It plugs into existing rec pipelines via lightweight event and model-output logging, then generates plain-language “why you saw this” explanations, bias/coverage diagnostics, and controllable knobs (e.g., diversity, novelty, fairness constraints) that can be A/B tested. The app focuses on practical governance: it flags risky patterns like filter bubbles, over-personalization, demographic skews, and sudden model drift after releases. It also creates audit-ready reports for internal reviews and regulators, including reproducible experiment snapshots. This is not a magic model that replaces your recommender; it’s an observability and control layer that makes your current system easier to trust, tune, and defend. Realistically, adoption requires engineering buy-in and clean data, so the product must be dead-simple to integrate and immediately useful within a week.