Search the Atlas

Search risks, controls, and glossary terms

HighMiscalibrationDemonstrated

Overconfidence & poor calibration

Model Accuracy & Reliability

Description

Output confidence is not calibrated to correctness, so users cannot tell reliable answers from unreliable ones.

Example scenario

A risk-rating assistant expresses high confidence on edge cases it actually gets wrong, driving bad approvals.

Real-world evidenceDemonstrated

Multiple published evaluations (MedQA, LegalBench, FinanceBench) have systematically demonstrated that general-purpose LLMs underperform domain specialists on regulated-domain tasks and often produce generic or overconfident responses. IBM Watson's oncology failures also partly reflected domain knowledge gaps. The risk is well-evidenced in research and field pilots but mass-harm production incidents specifically attributable to domain gaps alone are not cleanly documented.

Primary mitigations

  • Calibrated confidence/uncertainty estimates
  • abstention on low confidence
  • confidence display
  • selective prediction.

Detection signals

Calibration error (ECE); confidence-vs-accuracy curves; abstention-rate monitoring.

Mitigating controls

6
Non-agentic controls

Related risks in Model Accuracy & Reliability