Using Mathematical Models for Diagnostic Reasoning

“No doctor can be 100% sure that the diagnosis they’ve chosen is the right one. This is because medical diagnostics is always subject to incomplete and uncertain information,” says Dominik Battefeld, a research associate in subproject C05. Together with project lead Professor Dr. Stefan Kopp from the work group “Cognitive Systems and Social Interaction” at Bielefeld University, Battefeld created a mathematical model for errors made in diagnosing a seizure.

The model simulates the diagnostic procedure that medical professionals follow: the machine asks for information about the patient and orders various tests until it is confident enough to make a particular diagnosis. With their model, Battefeld and Kopp were thus able to systematically replicate diagnostic decision-making errors – errors that have also been demonstrated in psychological studies of physicians.

“Because the model can replicate typical errors, it helps to identify the sources of diagnostic mistakes,” says Kopp. “In the future, we would like to use this formal model to identify and resolve biased reasoning in physicians’ actual diagnostic records.”

Battefeld and Kopp published their findings in the article “Formalizing cognitive biases in medical diagnostic reasoning” and presented their work at the 8th Workshop on “Formal and Cognitive Reasoning,” held on September 19th, 2022, in Trier, Germany.

Link to the paper

Dominik Battefeld, researcher in project C05
Prof. Dr. Stefan Kopp, project leader of projects A01 and C05