Recommendation systems (and ranking algorithms within subscription aggregation systems) have an unfortunate quirk - they can suffer from their own success, and become too good at guessing what the user will find engaging. In this optimization cul-de-sac, users are left stuck in a rut and highly siloed from other users and other people. Platforms risk the user disengaging broadly from the ecosystem, and users risk hardening their preferences and losing their appreciation for novelty and challenge.
One way to counteract this would be to have an option constantly available to users to reset the data that platforms use to recommend them information (but ideally, not lose other pieces of their profile like personal connections, their own posts, etc). This kind of recommendation reset would offer the user the opportunity to explicitly decide to reset their relationship with the platform's recommendation system. That desire could be motivated by a wide variety of self-perceptions on a user's part - like feeling addicted to it, bored by it, or not feeling like the recommendations actually match their preferences.
In addition to putting more control into the user's hands, this kind of regulation would go a long way to address privacy concerns - the likes you made in your teens shouldn't be a part of your recommendation profile in your thirties.
Unlike many of the interventions on this site, which only really work as voluntary platform driven changes, this flavor of intervention is a perfect candidate for regulation, which could be framed in a couple of different ways: