[Paper Review] Machine learning and AI research for Patient Benefit: 20 Critical Questions on Transparency, Replicability, Ethics and Effectiveness

Sebastian J. Vollmer, Bilal A. Mateen+16 more|arXiv (Cornell University)|Dec 21, 2018

Artificial Intelligence in Healthcare and EducationMedicine33 references17 citations

TL;DR

This paper proposes 20 critical questions to enhance transparency, replicability, ethics, and effectiveness in machine learning and AI research for healthcare. The framework guides researchers through the entire project lifecycle—from data collection to implementation—aiming to establish an international consensus for responsible AI in health (AI-TREE), with the goal of improving patient outcomes through rigorous, accountable research practices.

ABSTRACT

Machine learning (ML), artificial intelligence (AI) and other modern statistical methods are providing new opportunities to operationalize previously untapped and rapidly growing sources of data for patient benefit. Whilst there is a lot of promising research currently being undertaken, the literature as a whole lacks: transparency; clear reporting to facilitate replicability; exploration for potential ethical concerns; and, clear demonstrations of effectiveness. There are many reasons for why these issues exist, but one of the most important that we provide a preliminary solution for here is the current lack of ML/AI- specific best practice guidance. Although there is no consensus on what best practice looks in this field, we believe that interdisciplinary groups pursuing research and impact projects in the ML/AI for health domain would benefit from answering a series of questions based on the important issues that exist when undertaking work of this nature. Here we present 20 questions that span the entire project life cycle, from inception, data analysis, and model evaluation, to implementation, as a means to facilitate project planning and post-hoc (structured) independent evaluation. By beginning to answer these questions in different settings, we can start to understand what constitutes a good answer, and we expect that the resulting discussion will be central to developing an international consensus framework for transparent, replicable, ethical and effective research in artificial intelligence (AI-TREE) for health.

Motivation & Objective

• Address the lack of transparency, replicability, ethical scrutiny, and demonstrated effectiveness in current ML/AI research for healthcare.
• Identify systemic gaps in reporting and methodology that hinder the translation of AI models into real-world clinical benefit.
• Provide a structured, interdisciplinary framework to guide ML/AI projects in health from inception to implementation.
• Foster the development of an international consensus standard for responsible AI in medicine (AI-TREE).
• Encourage post-hoc evaluation and project planning through a unified set of critical questions applicable across diverse settings.

Proposed method

• Develop a comprehensive list of 20 questions spanning the full project lifecycle of ML/AI in health research.
• Structure questions around core pillars: transparency, replicability, ethics, and effectiveness.
• Integrate interdisciplinary perspectives from clinicians, data scientists, ethicists, and health services researchers.
• Design questions to support both prospective project planning and retrospective evaluation of completed studies.
• Emphasize practical application by embedding questions at key stages: data sourcing, model development, validation, and implementation.
• Use the questions as a scaffold for improving reporting standards and enabling independent assessment of research quality.

Experimental results

Research questions

• RQ1 How can ML/AI research in healthcare be made more transparent and reproducible across diverse research teams and institutions?
• RQ2 What ethical risks emerge during the development and deployment of AI models in clinical settings, and how can they be proactively addressed?
• RQ3 To what extent do current ML/AI studies demonstrate real-world effectiveness and patient benefit beyond statistical performance?
• RQ4 How can interdisciplinary collaboration be structured to ensure that AI research in health is both scientifically robust and ethically sound?
• RQ5 What criteria define a 'good' answer to key methodological and ethical questions in AI for health, and how can consensus be built?

Key findings

• The 20-question framework provides a practical, actionable tool for improving the quality and accountability of ML/AI research in healthcare.
• The framework supports both project planning and independent evaluation, enhancing transparency and replicability across studies.
• By addressing ethical concerns early and systematically, the framework reduces the risk of harmful or biased AI applications in clinical settings.
• The questions are designed to be adaptable across different healthcare contexts, data types, and model development stages.
• The paper establishes a foundation for developing an international consensus standard—AI-TREE—for responsible AI in health research.
• The framework is intended to catalyze discussion and evolution of best practices in AI for patient benefit, with measurable impact on research quality and clinical translation.