A Multidimensional Framework for Classifying Human-AI Interactions in Clinical Trials

Summary

This paper proposes a multidimensional framework for identifying and comparing human-AI interactions (HAIIs) in clinical-trial records. The framework classifies each interaction by the AI task, the relationship between humans and AI, the interaction configuration, and the human groups involved. It defines HAII, reviews existing taxonomies, and extends them into a single approach designed for clinical-trial data. The authors purposively sampled 15 clinical trials from a previously reported dataset. Each trial was independently classified by two human reviewers and six large language model classifiers. The results indicate that LLMs could assist with systematic categorisation and synthesis of AI-related trials. However, the study also finds that human judgement remains important when trial records are incomplete or ambiguous. By making different forms of human involvement explicit, the framework is intended to support more consistent comparisons across studies and clearer analysis of how AI interventions are used in clinical trials.