Calibration
The degree to which a model's predicted probabilities match the actual likelihood of outcomes, ensuring that a model's confidence levels reflect its true accuracy.
In plain language
When a model says it is 90 percent confident, a well-calibrated model is right about 90 percent of the time. A poorly calibrated model might be right only 70 percent of the time while claiming 90 percent confidence.
Why this matters
Calibration directly impacts the quality of human decision-making that relies on AI recommendations. A well-calibrated model helps decision makers appropriately weight AI advice. Your governance framework should require calibration assessment as part of model validation before deployment.
Relevance
ImplementationCalibration is a measurable property assessed during model validation that determines how much weight decision makers should place on AI recommendations.
Related terms
Putting calibration into practice in your organisation?
Ready to transform your AI strategy?
Partner with Australia's AI strategy and governance specialists. From adoption roadmaps to ISO 42001 audit readiness.