Definition

Risk that AI/ML models perpetuate or amplify biases (e.g., under-reporting in certain populations, coding bias) affecting signal detection or case handling.

Example

Training data skewed toward certain regions or age groups leading to missed signals elsewhere.

Regulatory source

EMA reflection paper on AI, FDA AI/ML considerations

Common mistakes

Not assessing training data representativeness or demographic bias.

Inspector question

How do you assess and mitigate bias in AI tools used in PV?

Related terms