Human AI interactions captured as reusable data in notebooks
Human--AI Interactions as First-Class Data: A Formal Interaction Model and Runtime for Explainable AI Collaboration
Artificial IntelligenceHuman-Computer Interaction
Summary
Interactions between people and AI systems often disappear after a single use, making it hard to understand what happened or learn from past interactions. The authors propose treating these conversations and data exchanges as permanent, structured data to improve explainability and replay. They created a formal way to describe these interactions and built a tool for JupyterLab that saves and analyzes this data. This helps track how users behave and learn when working with AI assistants.
What this means in practice
- •For software engineers: Store and analyze detailed records of AI interactions within notebooks to improve debugging and collaboration workflows.
- •For learning platform developers: Capture and assess user interactions with AI tutors to better understand learner behaviors and tailor support.
Authors
Michela Taufer, Kin Ng, Ian Lumsden, Aashish Pandey, Rajasi Gore, Angelina Ju, Sukanya Moudgalya, Ewa Deelman, Sai Swaminathan
Abstract
AI assistants are increasingly embedded in computational notebooks, scientific workflows, software engineering tools, and learning systems, yet human-AI interactions are often treated as transient prompt-response exchanges rather than durable data products, hindering reproducibility, explainability, replay, behavioral analysis, and adaptive AI support. We argue that these interactions form a new data class requiring formal representation and introduce the Human-AI Interaction Description Language (HAIDL), a formal interaction model with a Backus-Naur Form-inspired concrete syntax. HAIDL encodes each interaction as a structured object capturing computational context, user request, prompt payload, AI response, feedback, problem context, provenance, and model metadata. We integrate this vision into HAIDL-X, a middleware for JupyterLab that instantiates HAIDL by capturing user-bot state, building contextualized prompts, interfacing with AI providers, collecting feedback, and storing interactions in a repository. We add a learning layer over HAIDL that maps interaction objects and notebook execution traces into reproducible measures of user behavior, including engagement, performance, help-seeking, error-oriented use, and learning personas. HAIDL-X provides a foundation for treating human-AI collaboration as structured, queryable, replayable, analyzable data.