Agentic data science with the Data Agent Kit
cloud.google.com·7h ago
AI Research
arXiv CS.AI·11d ago·5 min read
Defining and Categorising Human-AI Interactions in Clinical Trials: A Multidimensional Human-AI Classification Approach
TL;DR
A multidimensional framework for classifying human-AI interactions in clinical trials was developed, enhancing the consistency of identifying and comparing these interactions. The framework integrates AI tasks, human relationships, and interaction configurations, demonstrating the value of both human judgment and large language models in categorization.
✦ Why It Matters
Implement this framework to improve the categorization and analysis of AI interactions in your clinical trials today.
Key Takeaways
- Utilize the multidimensional framework for categorizing HAIIs in trials.
- Incorporate both human reviewers and AI classifiers for better accuracy.
- Focus on the four dimensions: AI tasks, relationships, configurations, and human groups.
- Ensure human judgment is applied in ambiguous cases for reliable categorization.
Full Summary
Human-AI interactions (HAIIs) in clinical trials are increasingly important as AI technologies are integrated into healthcare. This study introduces a multidimensional classification framework that categorizes HAIIs based on four dimensions: AI tasks, human-AI relationships, interaction configurations, and the groups of humans involved.
The researchers analyzed 15 clinical trials, using both human reviewers and six large language model (LLM) classifiers for categorization. The results showed that while LLMs can assist in categorization, human judgment remains crucial for interpreting incomplete or ambiguous trial records.
This framework aims to standardize the identification and comparison of HAIIs across clinical trials, ultimately supporting better synthesis of AI-related research in healthcare. The findings highlight the importance of structured categorization in understanding the diverse roles humans play in AI interventions.