A model can find a pattern in data and still draw the wrong conclusion. Domain knowledge—the specialized understanding of a field—helps people judge whether that pattern makes sense in the real situation. In AI projects, it complements data science and machine learning rather than being replaced by them. section 10.3 section 10.4.5
When a medical pattern misleads
A machine-learning model suggested that asthma reduced the risk associated with pneumonia. That conclusion was dangerous: patients with asthma faced substantially higher risk if they were not hospitalized promptly. Clinical expertise was needed to question the pattern, interpret the data in context, and check the model’s expected responses. section 10.3
Medical diagnosis shows why data alone is not enough. A clinician gathers records, observations, symptoms, and test results, then decides which details matter. Statistical correlations do not by themselves supply an understanding of disease mechanisms or that judgment. For consequential AI applications, involve domain experts in validating what a model has learned and how it will be used. section 7.0 section 10.3
