TUMuchSocialMedia

The Problem

You know it. You are opening instagram, tiktok, etc. to check a message a friend sent and 10 minutes later your still scrolling unconsciously through videos you won't remember longer than the blink of an eye. You spare a minute at the pedestrians crossing and automatically your thumb hovery over the purple icon.

What it does

Our solution doesn't focus on restricting your daily usage time or making it impossible to open apps. Most of those get deinstalled or avoided through loopholes anyways. We want to actively train your behavioural patterns and make you think about why you open an app and for how long you plan to do it.

The Agentic Part

When opening the app for the first time in a quick on-boarding users set their goals for which apps and at which times of the day they'd like to reduce the screen time. The App then blocks access to those apps and when trying to use them the user gets send to the App. There he has to convince the "Bouncer" Agent why he want to access the restricted apps. The Agent has a Memory of previous attempts, the users preferences and goals, and context like user calendar data, time, users screen time and more to estimate if he wants to allow the user access or not. Additionally we have a proactive agent that screens the users activity and can give reminders if he notices the user deviates from the argument he gave or looses himself to social media unintentionally.

How we built it

Frontend with Flutter, backend with Fastapi, supervision and reasoning agent with OpenAI API.

Challenges we ran into

Not starving. Regulating Pulses. Weighing up risk of short-term cardiovascular health decrease against long-term mental helath improvements.

Accomplishments that we're proud of

Specifying an initially very vague idea and making the final app many feautures we declared as future vision already.

What's next for TUMuchSocialMedia

Connection with other apps you want to substitute your social media use with (Duolingo, Anki, ...). Make the agent better learn the users usage patterns and adapting to behavioural changes.