Dopamine Maximization Blackhole

I recently worked on a project at CMU where our team designed a content recommendation engine — but one that understands the mental state of a user rather than just their click history.

The system asks initial mental health questions and then adjusts recommendations when it detects anxious or mindless scrolling behavior. The goal: improve mood, not maximize engagement.

What Platforms Actually Optimize For

TikTok, YouTube, and Instagram don't optimize for how you feel when you close the app. They optimize for how long you stay. Instagram's algorithm actively amplifies sexualized and violent content because it increases time on app. Joe Rogan and others have criticized the addictive nature of these timelines publicly — and the platforms have largely not changed.

The Business Opportunity

Here's what's interesting: this is also a competitive vulnerability. ChatGPT disrupted Google Search not by being louder but by being better for the user. A mentally sustainable social platform — one that actually improves your state instead of degrading it — could capture market share from the engagement-maximizing incumbents.

The better long-term retention numbers will come from users who feel good about how they spend time on your platform, not users who feel addicted to it. One of those is a moat. The other is a liability waiting for regulation to expose it.

The CMU project was a proof of concept. But the underlying idea — that recommendation systems could be designed around wellbeing rather than dopamine — isn't just an academic exercise.