HAIDL models human-AI notebook interactions as durable data
Original title: Human--AI Interactions as First-Class Data: A Formal Interaction Model and Runtime for Explainable AI Collaboration
HAIDL-X middleware for JupyterLab stores prompts, responses, feedback, and provenance and derives behavior measures from traces.
The paper argues that human-AI exchanges in notebooks, scientific workflows, software tools, and learning systems should be stored as durable data rather than transient prompts. It introduces HAIDL, a formal interaction model with a Backus-Naur Form-inspired syntax covering computational context, user request, prompt payload, AI response, feedback, problem context, provenance, and model metadata. HAIDL-X, a JupyterLab middleware, captures user-bot state, builds contextualized prompts, calls AI providers, collects feedback, and stores interactions. A learning layer maps those objects and notebook traces to measures of engagement, performance, help-seeking, error-oriented use, and learning personas.
- HAIDL encodes each exchange as a structured, provenance-bearing interaction object.
- HAIDL-X runs inside JupyterLab and stores interactions in a repository.
- The learning layer uses interaction objects and notebook execution traces.
- Reported measures include engagement, performance, help-seeking, and learning personas.
- The stated goal is replayable, queryable data for explainable collaboration.
From the original article
1 min read at arxiv.orgAI 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.
By Michela Taufer · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.