My academic work outside the operating room concerns the education of medical students and surgical residents, and in particular the place of artificial intelligence within it. I hold a faculty appointment at the American University of Beirut and previously directed the otolaryngology residency program at Sheikh Shakhbout Medical City in Abu Dhabi. The work falls into three areas: the critical appraisal of machine-generated information, the construction of assessment and study material, and simulation in surgical training.
Artificial intelligence and clinical reasoning
Large language models have become part of the way medical students read, revise and prepare for examinations. These systems produce fluent and confident text whose accuracy cannot be assumed. A model may support a recommendation with an authentic citation and still depart from accepted practice, and a learner without clinical experience has no means of detecting the discrepancy. I regard the appraisal of such output as a competency that must be taught explicitly, in the same manner as the appraisal of the published literature. My interest lies in the definition of that competency, in the methods by which it is taught, and in the measurement of its acquisition.
Assessment and study material
Examination questions and study material generated by a language model are of value only when each item can be traced to the evidence from which it was derived. I am interested in methods of generation that preserve this traceability, and in their application to formative assessment, where learners require a large volume of questions with immediate feedback.
Kalimat is the setting in which I apply these methods. It is a trilingual word game that I designed and published on the App Store, and its medical section is a question bank written for residents. Each item is a short clinical vignette that leads to a single answer, which may be a diagnosis, a drug, an operative technique, a complication, an anatomical structure, a classification, or a concept established by a guideline or a clinical trial. The answer is followed by an explanation of the reasoning and of the alternatives against which it was weighed, a statement of its clinical relevance, and the references on which it rests. The items are written in English, French and Arabic from the same body of evidence, drawn from current guidelines, consensus definitions and original trials, and I review each one before it is published.
The section offers a ranked daily challenge and an untimed study mode. It began with otolaryngology, pediatrics followed, and new items are added to each specialty every day.
Surgical simulation
Competence in otologic surgery is acquired through dissection of the human temporal bone, and cadaveric specimens are scarce and costly. I design and print anatomical models for surgical training. Printed temporal bones allow residents to rehearse the cortical mastoidectomy, and to learn the position of the structures at risk, before they dissect cadaveric bone. The same method is being extended to reconstructive surgery, with printed mandibles and cutting guides for the planning of fibular reconstruction. The process, from segmentation of the imaging to dissection of the printed model, is documented in a series of videos.