Fairness, calibration, and boundaries
Step 1 / 3·Rating bias and the role of calibration
0/3 steps done
Learning objectives — after this module you will:
- Recognize bias in ratings and calibration
- Use AI to support calibration without replacing the decision
- Keep transparency and the right to appeal
Rating bias and the role of calibration
Understand why calibration is needed and how AI helps.
Performance ratings are prone to bias: the halo effect, similarity bias, 'lenient' vs 'harsh' managers. CALIBRATION is a standardizing panel that reduces these skews. AI helps by aggregating the distribution (how many people at each level), flagging anomalies (a whole department at 5?), and comparing to a guideline curve — but the DECISION to adjust belongs to the panel, not the AI.
TTD's Performance Calibration tool
This tool lets you move ratings between levels and compare against a guideline curve — supporting the discussion. It does not automatically adjust anyone's score; the decision stays with the review panel.
Key takeaway: Calibration reduces rating bias; AI aggregates and flags anomalies, the panel decides.