Interactive Turing Test Game - Can you tell which response is from a human and which is from AI?
A gamified mental health awareness tool that challenges users to distinguish between human and AI-generated responses to mental health prompts.
This app is a playful, thought-provoking mental health awareness tool that gamifies Alan Turing's famous question — "Can machines think?" — in the context of lived mental health experiences. Users swipe through narratives in response to mental health prompts and try to guess whether they were written by a person with lived experience or a large language model (LLM).
This app supports Columbia’s University Mental Health Initiative by raising awareness, provoking reflection, and amplifying marginalized voices through accessible AI.
👉 https://turing-app.vercel.app
Deployed on Vercel with automatic deployments from the main branch.
Turing_APP/
├── pages/ # Next.js route pages
│ ├── index.js # Landing page with intro & start button
│ ├── game.js # Main swipe-based Turing Test experience
| ├── _app.js # App-level wrapper with context provider
│ └── about.js # App overview, mission, and credits
├── components/ # Reusable UI components
│ ├── Comments.js # User comment form and thread view
│ └── GameSettings.js # Toggle dark mode, font size, timer
├── contexts/
│ └── GameContext.js # Global state: settings, user inputs, comments
├── data/
│ └── turing_data.js # Prompt + human/AI answers with tags
├── styles/
│ └── globals.css # Global styles and dark mode
├── package.json # App dependencies and scripts
└── package-lock.json # Lock file for consistent builds
- turing-machine - Interactive implementation of the Turing Test
- vercel - Deployed and hosted on Vercel platform
- Next.js - React framework for web applications
- Mental Health - Awareness and education tool
- AI/ML - Human vs AI detection game
- Interactive Game - Swipe and click-based gameplay
- Two Game Modes:
- Swipe Mode: Swipe right for Human, left for AI (with visual indicators 🤖👤)
- Click Mode: Choose between two responses (randomized order)
- 15 randomized prompts per session
- Detailed results table showing question-by-question performance
- Dark mode toggle, font size slider, timer (30–90s/question)
- Human response identification tracking
- Users can filter prompts by mental health condition:
- Anxiety, Bipolar, Depression, OCD, PTSD, ADHD, Schizophrenia, etc.
- Users can leave comments per prompt, sharing thoughts and reactions
- Promotes reflection and dialogue around AI vs. human narrative tone
The turing_data.js file contains ~200+ paired responses sourced from:
- 🧍 Real people with lived mental health experience at Columbia (anonymous)
- 🤖 Popular LLMs like ChatGPT, Claude, Gemini, etc.
Each record includes:
prompthumanresponseairesponseconditiontag (e.g., Anxiety)typetag (e.g., Surprising, Helpful, Humanizing)
This project supports Columbia's Mental Health Initiative (MHI) and was inspired by:
- Ethical challenges in AI-generated MH content
- The Turing Test reimagined for public education and stigma reduction
- Elevating people with lived mental health experience (PWLEMH)
Full Proposal: “Human or AI? Using the Turing Test to Share and Deepen Perspectives on Mental Health”
- Fully responsive design
- Works on mobile, tablet, and desktop
- Dark mode and font size adjustments
- Keyboard support (← for AI, → for Human)
- Built by Zichen Zhao (Jackson)
- Columbia School of Social Work · Data Science Research Assistant
© 2025 Zichen Zhao. All rights reserved.
Use of this app is permitted for education, awareness, and research.
Redistribution or reuse of source code is prohibited without written permission.