AI and Medical Reasoning: Topic Collection in BMJ Digital Health & AI
A call for transparency in how artificial intelligence thinks in medicine
Why a Medical Reasoning Collection?
ARISE researchers Dr. Ethan Goh, Dr. Liam McCoy, and Dr. David Wu will collaborate with BMJ Digital Health & AI to guest-edit a new Topic Collection. Submissions will address the ongoing need for transparent insight into the reasoning processes of medical artificial intelligence.
Once confined to pattern recognition and automated task performance, advanced AI systems—particularly large language models (LLMs)—now demonstrate capacities previously reserved for human clinicians: the ability to reason through complex diagnostic and therapeutic challenges, collaborate meaningfully with healthcare providers, and even surpass human benchmarks on intricate medical reasoning tasks.
Recent landmark studies underline this transition. Randomized trials reveal that physician performance improves markedly when supported by AI reasoning assistants (Goh et al., Nature Medicine, 2025). Benchmarks indicate expert-level medical question answering capabilities (Singhal et al., Nature Medicine, 2025). Moreover, novel evaluations suggest LLMs achieve superhuman outcomes on physician-level diagnostic tasks (Brodeur et al., arXiv, 2024).
Yet, while the quality of final outputs from these systems inspires optimism, we remain largely blind to how exactly these models reach their conclusions.
Current evaluation methods predominantly score AI systems by their final decisions alone, treating internal reasoning processes as opaque. This “black box” approach leaves critical gaps in our understanding of AI’s strengths and vulnerabilities—especially in the nuanced, high-stakes environments of clinical decision-making.
Context and Call for Submissions
The ARISE collaborators’ guest-edited Topic Collection, “Decoding Clinical Reasoning in AI Systems,” in BMJ Digital Health & AI seeks to address this challenge head-on. We invite submissions that not only assess performance but critically interrogate and elucidate the reasoning processes of clinical AI. By doing so, we aim to promote transparency, accountability, and trustworthiness—qualities essential for the ethical integration of AI into healthcare.
We encourage contributions from diverse disciplinary perspectives, including cognitive science, informatics, philosophy, health law, and beyond. Global submissions, especially from underrepresented regions, are explicitly invited.
This initiative aspires to shift the conversation from “How accurate is AI?” toward “How does AI reason—and how can we trust it?” Together, we can build the foundations of a more transparent, accountable, and clinically impactful AI.
Submission Details
Submission Deadline: January 16th, 2026
For submission details, author guidelines, and APC discounts, visit our Topic Collection Page.
Guest Editors:
Ethan Goh, MD, MS – Stanford University
Liam G. McCoy, MD, MSc – University of Alberta
David Wu, MD, PhD – Mass General Brigham
Contact: topic.collections@bmj.com
