AImposter

AImposter Game
Gameplay
Gameplay
Login Screen
Account Creation
Chatroom

IMPORTANT NOTE

due to wifi outage, we are unable to deploy and upload our video. our product is fully working, and the current video is just a test upload. mockup link: https://www.figma.com/design/vMmwDQAqQmx2TF7pA9l9LK/hackathon?t=9X5o76W69rQoqEXj-1

Inspiration

We were inspired by Odd One Out videos, where a group tries to figure out who doesn’t belong, but we wanted to bring that concept into the world of AI. At the same time, we noticed how artificial intelligence has become part of our everyday conversations, from study help to emotional support. That led us to ask: What if an AI joined your group chat and no one knew who it was?

From that idea came AI-mposter: a platform that combines social deduction with collaborative AI chatrooms, blurring the line between human and machine interaction.

How we built it

We built AI-mposter as a real-time web platform featuring two main experiences: Multi-user AI Chatrooms: one AI and multiple real users can collaborate in themed rooms such as study sessions, classroom discussions, supportive group chats, or fun. AI Imposter Game: inspired by Odd One Out and social deduction games, players must chat, observe, and vote to uncover which participant is secretly an AI.

Under the hood, we used: Flask for backend routing and session handling Socket.IO for real-time chat communication Databases for user management and chatroom storage Figma for UI/UX design JWL encryption to protect user data and messages

Challenges we ran into

Synchronizing chat data across multiple clients and the AI in real time Preventing AI responses from feeling repetitive or robotic Ensuring stable Socket.IO connections under multiple concurrent users Designing a UI that made the game engaging yet accessible Managing state across multiple rooms and game sessions

Accomplishments that we're proud of

Built a fully functional multi-user chat system that supports real-time AI and human interaction Designed an original AI-driven game mechanic inspired by Odd One Out and social deduction games Created a smooth, modern interface that makes chatting and gameplay feel natural and fun Balanced technical complexity with human psychology, exploring how people detect or trust AI in group settings Learned to integrate multiple technologies (Flask, Socket.IO, encryption, databases) under tight hackathon time pressure

What we learned

Managing real-time, multi-user communication efficiently Calibrating AI tone and behavior for different roles (teacher, study buddy, or imposter) Exploring trust and deception dynamics in AI-human conversations Building secure systems that handle sensitive data responsibly

What's next for AImposter

Expanding AI roles for more personality, expertise, and interactive behavior Improving AI imposter mechanics to be more sneaky Analytics and feedback for improved responses and actions And more !!