Medical Humanities

Share               

Medical Artificial Intelligence

Melvin Chen


Given that we are currently in the age of the Fourth Industrial Revolution, there is a burgeoning interest in programmatic research on medical AI. In the field of the biomedical sciences, data-driven drug development, robot-assisted surgery, and automated frontline healthcare are just some of the possibilities that will drive the quest for medical AI. With respect to frontline healthcare, the question we must ask ourselves is whether diagnosis, provision of care, and doctor-patient and nurse-patient relations can be automated. In order for an effective diagnosis to be made, doctors must reason about causal claims that relate putative causes to the observed effects (the patient’s symptoms). There is no doubt that AI programs typically exceed human beings in raw computational power and can handle large amounts of data better than even the most competent medical professionals. The best AI programs and machines in the market, however, rely on statistical correlation and curve-fitting, which are decent guides to causation but no guarantee of the latter. Judea Pearl has recently provided a formalized notion of causation (his SCM or Structural Causal Model, which relies on the use of diagrams and equation models). Can the powers of the causal revolution (as begun by Pearl) be harnessed and can this formalized notion of causation be used to yield medical AI systems that make causal inferences as competently as human doctors do (often with the data being underdetermined or noisy and with the web of causation being complex)? Can medical systems be designed to imagine about possible worlds in order to reason counterfactually, with a view to providing the best possible diagnoses? This is the first challenge for imagination machines in medical AI. In addition, medical professionals do not merely treat the medical symptoms but also have to provide care for the patient. Can medical AI systems be sufficiently endowed with a recreative imagination that will allow them to think from the perspective of the patient, make sense of the narrative of that patient, and provide the best possible care? This is the second challenge for imagination machines in medical AI. This work-in-progress on medical AI is a collaborative project between Dr. Melvin Chen (Philosophy) and Assoc. Prof. Chew Lock Yue (Physics).​