HomeLearnCoursesHackathonsAccount
Recommender Systems
Implicit vs. Explicit Feedback · 1/2

Two kinds of signal about what people want

Explicit feedback is when a user directly tells the system their preference: a five-star rating, a thumbs up or down, an explicit 'not interested' click. It's clean, easy to interpret, and unambiguous, a one-star rating clearly means the user disliked something. The problem is that explicit feedback is rare. Most users never bother to rate anything, so systems that rely solely on explicit feedback are working with a tiny, self-selected slice of interactions, and that slice is often biased toward people with strong opinions in either direction.

Implicit feedback is inferred from behavior rather than stated directly: clicks, watch time, purchases, skips, how far someone scrolled, whether they replayed something. It is abundant, since it's generated automatically by ordinary usage rather than requiring users to do extra work, but it's also far noisier and harder to interpret. A click doesn't necessarily mean interest, it might mean curiosity, misclick, or clickbait. A short watch time might mean dislike, or it might mean the video was simply short. Implicit signals are ambiguous in a way explicit ratings aren't.