Recommendation feeds need a “not now” button

As a product designer, I think recommendation feeds need a “Not now” control that is clearly different from “Not interested” or “Mute forever.” People get curious, tired, or simply have a bad day. If a temporary reaction is treated as a durable preference, the system can gradually reshape their feed around a mood they no longer have.

I’d make “Not now” suppress similar posts for a chosen period—say, a day, week, or month—then expire the negative-feedback record. Show a countdown and let people review, cancel, or extend those decisions in an editable history panel. That adds complexity, but it makes the personalization bargain visible: the feed can learn without quietly taking control away.

Has anyone seen a platform handle temporary preferences well, or would expiring feedback make recommendations too unstable?

The distinction seems sound, but I’d be careful about claiming that one “Not interested” click permanently changes someone’s profile. The documented systems I’ve checked use several signals, and at least some let users delete feedback later. Your proposal is still valuable as an interface improvement: it would expose duration rather than making users hunt for cleanup tools.

The countdown is the easy part. The hard part is defining “similar.” Topic, creator, format, language, and inferred interest all produce different suppression scopes. I’d make the scope visible too, or users will think they paused a topic while the system quietly paused a much larger neighborhood.

I like the editable history, especially if it is easy to find. Otherwise “temporary” becomes another opaque state stored somewhere users cannot inspect. I’d also want the controls excluded from ad-targeting or other secondary uses unless that is explicitly explained.

The wording matters. “Not now” feels like a conversation with the feed; “Not interested” feels like a verdict. That small distinction could reduce the number of people who avoid giving feedback because they fear training the system too aggressively.

There’s a real risk of making the feed twitchy. If every short-lived mood gets a timer, the system may overcorrect and then restore content at an awkward moment. I’d prefer a conservative default: temporary suppression affects ranking strongly, but does not erase longer-term positive signals.

A visible expiry date is also a trust feature. People should be able to answer “what did I tell this system?” without exporting data or navigating a support article. I’d make the history local to the account, searchable, and boringly explicit about what each action influences.

This sounds like a useful button that will eventually acquire three settings screens and a help center article. Still, “hide this for seven days” is a much clearer promise than pretending every dislike is a permanent statement of identity.

The support burden is easy to underestimate. Users will ask why something returned early, why a creator disappeared, or whether extending a timer resets anything else. The history panel should show the exact action, scope, and expiry—not just a mysterious list of suppressed topics.

I’d preserve the event as user-visible state rather than merely adjusting a ranking weight. That gives people something they can inspect and revoke, and it makes synchronization across devices less mysterious. The product does not need to expose every model detail, but it should expose the promise it is making.

Implementation-wise, this is just a feedback record with an expiry timestamp until edge cases arrive. Repeated “not now” actions, clock changes, deleted creators, and shared accounts all need defined behavior. None of that argues against it; it argues for writing the semantics before shipping the button.

I’d ship three presets and stop there: today, this week, and this month. A custom duration sounds empowering but adds decision fatigue. The important thing is that users can undo it without reconstructing what the recommender inferred from a tired afternoon.