18.3 Responsible AI: Detecting & Monitoring Bias

Key Takeaways

  • Agent Platform data bias metrics include Difference in Population Size and Difference in Positive Proportions in True Labels, computed before training.
  • Agent Platform model bias metrics include Accuracy Difference, Difference in Positive Proportions in Predicted Labels, Recall Difference, Specificity Difference, and Difference in Ratio of Error Types.
  • In Agent Platform's fairness metrics, a positive value generally indicates potential bias favoring slice 1, zero indicates no difference, and a negative value favors slice 2.
  • Bias can enter through unrepresentative data, historical labels, proxy features, and deployment context, so fairness checks belong before training, before deployment, and in production monitoring.
  • For gen AI, responsible AI practice combines safety filters, evaluation of outputs across user groups, and human review of high-impact use cases.
Last updated: September 2026

The exam guide lists responsible AI practices (for example, monitoring for bias) as part of identifying risks to AI solutions. The exam expects you to recognize bias sources, pick suitable fairness metrics, and put checks into the lifecycle, not just to state principles.

Where Bias Comes From

SourceExample
RepresentationA smile-detection dataset with far fewer older people
Historical labelsPast hiring decisions that favored one group become the "ground truth"
MeasurementDifferent data quality across regions (older devices, missing fields)
Proxy featuresZIP code standing in for protected attributes
AggregationOne model for populations that behave differently
Deployment contextA model used for decisions it wasn't designed or validated for
Feedback loopsA model's decisions shape future training data, such as fewer loans giving fewer repayment examples

Agent Platform Fairness Metrics

Agent Platform provides bias metrics (preview) comparing slice 1 (the group measured) against slice 2 (the comparison group). For most metrics, positive values suggest potential bias favoring slice 1, zero means no difference, and negative values favor slice 2. The evaluation pipeline components DetectDataBiasOp and DetectModelBiasOp compute them.

Data bias metrics (before training)

Example: 200 California applicants (60 accepted) and 100 Florida applicants (20 accepted).

MetricFormulaExample
Difference in Population Size(n₁ − n₂) / (n₁ + n₂)(200 − 100) / 300 = 0.33
Difference in Positive Proportions in True Labels (DPPTL), also called label imbalancepositives₁/n₁ − positives₂/n₂60/200 − 20/100 = 0.10

Model bias metrics (after training)

Confusion matrices: California TP 50, FN 10, FP 20, TN 120. Florida TP 20, FN 0, FP 30, TN 50.

MetricComparesExample
Accuracy Difference(TP+TN)/n for each slice170/200 − 70/100 = 0.15
Difference in Positive Proportions in Predicted Labels (DPPPL)Predicted positives / n70/200 − 50/100 = −0.15
Recall DifferenceTP / (TP+FN)50/60 − 20/20 = −0.17
Specificity DifferenceTN / (TN+FP)120/140 − 50/80 ≈ 0.23
Difference in Ratio of Error Types (treatment equality)FN/FP for each slice10/20 − 0/30 = 0.50. Both slices have 30 errors, but California's errors include missed acceptances while Florida's are all false acceptances

Reading the example: the model is more accurate for California (+0.15) but predicts acceptance more often for Florida (−0.15) and has higher recall for Florida (−0.17). Which gap matters depends on the harm. In admissions, wrongly rejecting qualified applicants (recall) may be the main concern.

Choosing Fairness Criteria

There is no single fairness metric. They can conflict, so choose based on the decision's harms:

If the main harm is...Watch
Qualified people being denied (missed opportunities)Recall difference (equal opportunity)
Unqualified people being wrongly approved or flaggedFalse positive rate or specificity difference
Unequal selection rates across groupsDPPPL (demographic parity)
Overall quality gapsAccuracy difference plus per-slice precision and recall

Involve domain, legal, and affected stakeholders in picking metrics and acceptable thresholds.

Putting Fairness into the Lifecycle

StagePractice
Problem framingDecide if ML is appropriate. Document intended use and misuse
DataCheck representation (Difference in Population Size) and label imbalance (DPPTL). Look for proxies
TrainingReweight or resample underrepresented groups. Consider constraints or separate thresholds where lawful
EvaluationSlice metrics and bias metrics as pipeline gates (Chapter 15)
DeploymentCanary monitoring by segment. Human review for high-impact decisions
ProductionMonitor prediction distributions and outcomes by slice, and retrain or roll back when gaps widen
DocumentationRecord data sources, evaluation slices, limitations, and approvals with the model version

Mitigation Techniques

  • Collect more representative data for underrepresented groups. This is often the most effective fix.
  • Reweight or resample training examples, for example with AutoML's row weight column.
  • Remove or transform proxy features, and test whether performance gaps shrink.
  • Adjust decision thresholds per operating context where policy and law allow.
  • Human-in-the-loop review for borderline or high-impact predictions.
  • Explain predictions (Section 18.4) to find features driving unequal outcomes.

Monitoring Fairness in Production

Fairness at launch doesn't guarantee fairness later, because populations and behavior change.

  1. Log the slice attributes needed for analysis (where lawful and privacy-approved) alongside predictions.
  2. Schedule bias metrics on recent predictions and, once labels arrive, on outcomes. For example, run the model bias component monthly in a pipeline.
  3. Alert on thresholds agreed with stakeholders, such as a recall difference beyond ±0.05.
  4. Investigate root causes with slice-level feature distributions and explanations before retraining.
  5. Record decisions (accepted trade-offs, mitigations) with the model version for accountability.

Responsible Gen AI

  • Safety filters and Model Armor reduce harmful outputs (Section 18.2).
  • Evaluate across user groups and languages, because quality and safety can vary by dialect or topic (Gen AI evals, Chapter 7).
  • Grounding and citations reduce fabricated claims that can mislead users.
  • Transparency: tell users they are interacting with AI and where content is AI-generated.
  • Human oversight for medical, legal, financial, and employment uses.
Test Your Knowledge

Before training a hiring model, a team finds that 90% of historical 'hired' labels belong to one group, even though applicants were evenly split. Which Agent Platform metric directly quantifies this label imbalance?

A
B
C
D
Test Your Knowledge

For slice 1, TP = 40 and FN = 10. For slice 2, TP = 45 and FN = 5. What is the Recall Difference (slice 1 minus slice 2), and what does its sign suggest?

A
B
C
D
Test Your Knowledge

A lending model's overall AUC improved in a new version, but approval recall for one protected group dropped substantially. What should the deployment pipeline do?

A
B
C
D