CAT-AI: Supporting Teacher Workflows with AI-Assisted Exercise Creation

CAT-AI is a web-based system that supports K-12 teachers in creating, customizing, and exporting educational exercises through AI assistance. Rather than replacing teachers' expertise, CAT-AI operationalizes their natural authoring workflow across three integrated phases:

• Conceptualization — Teachers specify pedagogical parameters (subject, difficulty, exercise type) through structured fields instead of open-ended prompts, eliminating the need for prompt engineering.
• Transformation — A unified WYSIWYG editor enables both direct text manipulation and targeted AI-assisted regeneration, with confidence indicators flagging potentially unreliable content.
• Finalization — The system handles export formatting automatically, producing classroom-ready materials without requiring external tools.

Live demo: https://catai.paperbackwriters.club/

Prerequisites

• A valid OpenAI API Key
• Node.js (recommended: latest LTS version)
• npm (comes with Node.js)
• Optional: nodemon for automatic server restarts

---

Getting Started

1: OpenAI API Key Setup

Create a .env file inside the /server directory with your API key:

1. Create a file named .env in the server folder.
2. Add the following line: OPENAI_API_KEY="your_openai_api_key_here"

2: Start the Client

cd client
npm install
npm run dev

3: Start the Server

cd server
npm install
nodemon index.mjs  # or node index.mjs

If nodemon is not installed globally, you can install it with:

npm install -g nodemon

4: Open CAT-AI

Open http://localhost:5173 in your browser.

---

Citation

If you use this work, please cite:

@inproceedings{calo2026catai,
  author    = {Calo, Tommaso and Cuccu, Lorenzo and De Russis, Luigi},
  title     = {{CAT-AI}: Supporting Teacher Workflows with {AI}-Assisted Exercise Creation},
  booktitle = {Extended Abstracts of the 2026 CHI Conference on Human Factors in Computing Systems},
  series    = {CHI EA '26},
  year      = {2026},
  isbn      = {979-8-4007-2281-3/2026/04},
  location  = {Barcelona, Spain},
  publisher = {ACM},
  address   = {New York, NY, USA},
  doi       = {10.1145/3772363.3799367},
}