PAI Chatbot

Accessibility-teaching AI

PAI chatbot is a MERN-stack prototype which was built for a lab study as part of my bachelor thesis. It's a chatbot which uses an Anthropic API key to query their large language models, and was built to act as a coding-assistant which facilitates learning about accessibility practices in web development. Features include a system prompt containing instructions to intentionally leave gaps in generated code, such as an empty alternative text for an image, and nudging developers towards filling them, as well as contextually aware prompt suggestions which serve to help users carry on conversations in meaningful ways or otherwise alleviate the cognitive load of formulating prompts to bridge the gap for novice users who are not knowledgeable enough about accessibility to ask the right questions. Conversations between participants and PAI in the lab study are stored in a Mongo Database for research purposes.