Link
You can play the game online at the following link: 2 Truths and AI Game.
Project Overview
“2 Truths and AI” reimagines the classic party game “Two Truths and a Lie” by introducing AI-generated lies to reduce performance pressure and lower barriers to participation. The game creates situations where people can naturally talk and learn about each other while maintaining the fun and engagement of social games.
Classic icebreaker games rely heavily on players’ creativity and social confidence to invent convincing lies, which can be intimidating for many participants. Our project bridges this gap by using generative AI to incentivize group conversation while reducing the awkwardness of traditional icebreakers. The single-device voting mechanic leverages herd behavior and promotes democratic participation, keeping everyone engaged in the same conversation.
Key Features
- AI-Generated Lies: Multiple Azure GPT models create convincing lies based on players’ truths
- Role-Based Gameplay: “Rainbow Chameleon” role keeps all players engaged throughout every round
- Single-Device Design: Intentional choice to avoid sync issues and promote group discussion
- Dynamic Feedback: Clear visual indicators, animations, and color coding for all interactions
- Customizable Settings: Player count, round limits, timer duration, and AI model selection
- Game Statistics: Detailed metrics like “Fastest Guesser” and “Trickiest Chameleon” with export functionality
Iterative Design Process
The project was developed through three rounds of user research with our target demographic of party-game players:
User Needs Identification: Structured interviews explored experiences with icebreakers and attitudes toward AI-generated content in games, establishing core requirements and design direction.
Low-Fidelity Evaluation: Participants reviewed mockups using think-aloud protocols, revealing issues with text visibility, navigation clarity, and the need for better AI model explanations.
Functional Testing: Small groups played full games, providing feedback that led to simplified instructions, enhanced tooltips, improved AI prompt engineering, and better visual differentiation.
Key improvements included increasing font sizes, moving navigation buttons to prominent positions, adding hover tooltips with video demos, implementing “Pass the Device” modals for privacy, and refining color coding for better feedback. The iterative process resulted in a polished, accessible party game that successfully reduces social pressure while maintaining engagement.
User Impact
Players consistently expressed enjoyment with the game. The AI-generated lies created genuine uncertainty and debate, successfully sparking the natural social conversations the game was designed to promote. The iterative design process, informed by continuous user feedback, resulted in an experience that is more accessible and less awkward than traditional icebreakers while preserving their personal relevance.
Tools & Technologies
- AI/ML: Microsoft Azure GPT Models (gpt-35-turbo, gpt-5-nano, gpt-5-mini, gpt-4o, gpt-4.1-mini)
- Frontend: React, JavaScript/TypeScript
- Design: Figma
- Development: Visual Studio Code, WSL/Linux
- Storage: Browser sessionStorage API