Discussion Paper: Social accountability for students in a machine learning era

Abstract

Over the last 30 years, there have been repeated calls to integrate health informatics into undergraduate health professional curricula, in recognition of the integral role computing plays in medicine. The rise of big data sets in health, and the application of advanced computer algorithms to interrogate these, is yet another call for health professionals to receive appropriate training in these technologies. Machine learning (ML) algorithms can learn tasks or make decisions without a requirement for specific behaviours to be pre-programmed. High-impact literature has described ML approaches to clinical problems such as achieving more accurate and timely diagnoses, increasing precision of prognosis and guiding treatment. Despite the promise of ML in healthcare, there are risks of adverse outcomes, unanticipated consequences, misuse and even abuse of ML technologies. For health professionals to advocate for patients and hold those developing ML algorithms in healthcare accountable, they must feel comfortable discussing the fundamental concepts and limitations of ML in healthcare. Healthcare professionals are uniquely positioned to identify problems that could be solved by ML and related technologies. Yet, there is inadequate coverage of ML, or of the wider field of health informatics, in most medical curricula. To create future health professionals who can advocate for positive change and ensure that patients remain at the centre of ML applications in healthcare, we must provide future health professionals with an understanding of how ML will change healthcare delivery and the doctor–patient dynamic, as well as new ethical challenges that arise with the digital healthcare revolution.