Getting Started with LUCID
LUCID is a lightweight, extensible research platform that enables researchers to embed GPT-powered chatbot interactions inside Qualtrics surveys. This guide walks you through setup in just a few steps.
Step 1: Deploy Your Own GPT Server
To keep your OpenAI API key secure, LUCID requires each researcher to deploy their own backend server. This prevents survey participants from ever seeing or accessing your API key, which could otherwise happen if it were embedded directly in Qualtrics JavaScript.
The LUCID backend README walks you through:
- Creating an OpenAI API key
- Registering a GitHub account
- Deploying the LUCID backend to Vercel (one-click setup)
After setup, you’ll have your own private chatbot server, ready to connect with Qualtrics.
Step 2: Choose a Template
Head to the Templates Page to download pre-built .QSF files. Each template demonstrates a different type of human-AI interaction:
- Starter Template (single back-and-forth conversation with a GPT-powered chatbot)
- Between-subject Manipulation in Conversation (randomly allocate participants to one of two GPT-powered chatbots)
Step 3: Modifying Our Templates to Run Your Study
You'll then need to customize our templates for your own needs. At this point, following our Configuration Guide to:
- Edit the embedded URL for your deployed GPT endpoint
- Customize prompt wording if needed
- Launch your survey!
Using LUCID in your research?
Garvey, Aaron G. and Simon J. Blanchard (2025). Generative AI as a Research Confederate: The LUCID Methodological Framework and Toolkit for Controlled Human-AI Interactions in Survey Research. Working Paper at SSRN.