Explainer
What Is Accuracy Equality?
Equal accuracy across groups can hide two completely different, offsetting error profiles.
diff --git a/explainers/accuracy-not-enough-healthcare-ai.html b/explainers/accuracy-not-enough-healthcare-ai.html index 80e4e1a5..d0eb712b 100644 --- a/explainers/accuracy-not-enough-healthcare-ai.html +++ b/explainers/accuracy-not-enough-healthcare-ai.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Why Accuracy Is Not Enough in Healthcare AI
A 95% accurate model can still miss the sickest patients in one group.
diff --git a/explainers/adversarial-debiasing.html b/explainers/adversarial-debiasing.html index 994a95bc..db808b02 100644 --- a/explainers/adversarial-debiasing.html +++ b/explainers/adversarial-debiasing.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Adversarial Debiasing
Train a predictor against a second model that tries to read the group from its output.
diff --git a/explainers/adverse-action-notices.html b/explainers/adverse-action-notices.html index 1fe1fc30..2f2ef200 100644 --- a/explainers/adverse-action-notices.html +++ b/explainers/adverse-action-notices.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Adverse Action Notices
The reasons a lender must give are the one place a model's inputs are forced into the open.
diff --git a/explainers/aggregation-bias.html b/explainers/aggregation-bias.html index 3a230c66..a59d6b9f 100644 --- a/explainers/aggregation-bias.html +++ b/explainers/aggregation-bias.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Aggregation Bias
One model for everyone assumes one relationship.
diff --git a/explainers/ai-hallucinations.html b/explainers/ai-hallucinations.html index 67483f7f..13b90157 100644 --- a/explainers/ai-hallucinations.html +++ b/explainers/ai-hallucinations.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
AI Hallucinations
Confident outputs that are still wrong.
diff --git a/explainers/ai-objectivity-myth.html b/explainers/ai-objectivity-myth.html index 32554a30..b96e18f7 100644 --- a/explainers/ai-objectivity-myth.html +++ b/explainers/ai-objectivity-myth.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
The Biggest Myth About AI Objectivity
"It's just math" is not a defense
diff --git a/explainers/algorithmic-impact-assessment.html b/explainers/algorithmic-impact-assessment.html index fb3c62df..25b43660 100644 --- a/explainers/algorithmic-impact-assessment.html +++ b/explainers/algorithmic-impact-assessment.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Algorithmic Impact Assessment
Who does the decision affect, and how do you know?
diff --git a/explainers/algorithmic-recourse.html b/explainers/algorithmic-recourse.html index e61bd3f6..0c20535e 100644 --- a/explainers/algorithmic-recourse.html +++ b/explainers/algorithmic-recourse.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Algorithmic Recourse
What you can actually change to flip a model's decision.
diff --git a/explainers/auc-gaps-by-group.html b/explainers/auc-gaps-by-group.html index 29dab8af..0c5de23f 100644 --- a/explainers/auc-gaps-by-group.html +++ b/explainers/auc-gaps-by-group.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
What Is an AUC Gap Between Groups?
Equal ranking quality can hide unequal errors, and a threshold can hide a ranking gap.
diff --git a/explainers/automation-bias.html b/explainers/automation-bias.html index 6adb1a20..388a8356 100644 --- a/explainers/automation-bias.html +++ b/explainers/automation-bias.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
What Is Automation Bias?
When humans defer to algorithms, bias gets automated too.
diff --git a/explainers/base-rate-fallacy.html b/explainers/base-rate-fallacy.html index e3384b5e..4c5e27af 100644 --- a/explainers/base-rate-fallacy.html +++ b/explainers/base-rate-fallacy.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
What Is the Base Rate Fallacy?
Why ignoring background prevalence makes screening tools mostly wrong - and drives fairness metric conflicts.
diff --git a/explainers/bias-variance-tradeoff.html b/explainers/bias-variance-tradeoff.html index 51bf2596..60a943b1 100644 --- a/explainers/bias-variance-tradeoff.html +++ b/explainers/bias-variance-tradeoff.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
What Is the Bias-Variance Trade-off?
Why an overfit model can memorize the majority and fail the minority.
diff --git a/explainers/bootstrap-confidence-intervals.html b/explainers/bootstrap-confidence-intervals.html index f1456cc3..e0e5dae5 100644 --- a/explainers/bootstrap-confidence-intervals.html +++ b/explainers/bootstrap-confidence-intervals.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Bootstrap Confidence Intervals and Permutation Tests
A fairness gap without an interval attached is a number, not a finding.
diff --git a/explainers/calibration.html b/explainers/calibration.html index 12551b33..b9bb1b09 100644 --- a/explainers/calibration.html +++ b/explainers/calibration.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Calibration
When a model score actually means what it says.
diff --git a/explainers/causal-graphs-for-fairness.html b/explainers/causal-graphs-for-fairness.html index cbee5ba4..e7e73b42 100644 --- a/explainers/causal-graphs-for-fairness.html +++ b/explainers/causal-graphs-for-fairness.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Causal Graphs for Fairness
Two analysts, same data, opposite verdicts: the difference is a drawing.
diff --git a/explainers/choosing-a-fairness-metric.html b/explainers/choosing-a-fairness-metric.html index 74501a04..6a4bab92 100644 --- a/explainers/choosing-a-fairness-metric.html +++ b/explainers/choosing-a-fairness-metric.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Choosing a Fairness Metric
Start from the harm, not from the formula.
diff --git a/explainers/choosing-a-mitigation-strategy.html b/explainers/choosing-a-mitigation-strategy.html index 15777e92..c252af9f 100644 --- a/explainers/choosing-a-mitigation-strategy.html +++ b/explainers/choosing-a-mitigation-strategy.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Choosing a Mitigation Strategy
What each of the five strategies needs, costs and leaves.
diff --git a/explainers/class-imbalance.html b/explainers/class-imbalance.html index 75705772..e7aef603 100644 --- a/explainers/class-imbalance.html +++ b/explainers/class-imbalance.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
What Is Class Imbalance?
When a 99% accuracy score just means the model ignored the 1% that mattered.
diff --git a/explainers/clinical-score-miscalibration.html b/explainers/clinical-score-miscalibration.html index 1c6a4e25..67364b1e 100644 --- a/explainers/clinical-score-miscalibration.html +++ b/explainers/clinical-score-miscalibration.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Miscalibration in Clinical Risk Scores Across Groups
The same risk score can mean a different real-world risk depending on the patient's group.
diff --git a/explainers/compounding-disparities.html b/explainers/compounding-disparities.html index b7e1a8c7..7e2752da 100644 --- a/explainers/compounding-disparities.html +++ b/explainers/compounding-disparities.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Compounding Disparities
Four stages at 0.9 are 0.66 end to end.
diff --git a/explainers/conditional-demographic-parity.html b/explainers/conditional-demographic-parity.html index 6ed38ef3..70c64db7 100644 --- a/explainers/conditional-demographic-parity.html +++ b/explainers/conditional-demographic-parity.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
What Is Conditional Demographic Parity?
Checking parity within strata of a chosen legitimate factor, not across the whole population.
diff --git a/explainers/confidence-intervals-for-ratios.html b/explainers/confidence-intervals-for-ratios.html index 610b0520..6b2a59bf 100644 --- a/explainers/confidence-intervals-for-ratios.html +++ b/explainers/confidence-intervals-for-ratios.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Confidence Intervals for Ratios
A ratio of two noisy rates is skewed, and with a near-zero denominator it cannot be estimated.
diff --git a/explainers/confounding-variable.html b/explainers/confounding-variable.html index 645faf21..e268caf8 100644 --- a/explainers/confounding-variable.html +++ b/explainers/confounding-variable.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Confounding Variable
When a hidden cause makes two things look connected.
diff --git a/explainers/confusion-matrix.html b/explainers/confusion-matrix.html index 0ef6af39..4831904f 100644 --- a/explainers/confusion-matrix.html +++ b/explainers/confusion-matrix.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
What Is a Confusion Matrix?
The foundational building block behind most fairness metrics.
diff --git a/explainers/cost-sensitive-evaluation.html b/explainers/cost-sensitive-evaluation.html index e257acfb..9cf36827 100644 --- a/explainers/cost-sensitive-evaluation.html +++ b/explainers/cost-sensitive-evaluation.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
How Does Unequal Cost Change a Fairness Evaluation?
A false positive and a false negative rarely cost the same, and can cost differently by group.
diff --git a/explainers/counterfactual-explanation.html b/explainers/counterfactual-explanation.html index d3efa7db..d6b645f7 100644 --- a/explainers/counterfactual-explanation.html +++ b/explainers/counterfactual-explanation.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
What Is a Counterfactual Explanation?
One asks what changed the outcome. The other asks whether the outcome should have changed at all.
diff --git a/explainers/counterfactual-fairness.html b/explainers/counterfactual-fairness.html index edd2742b..0ab636f8 100644 --- a/explainers/counterfactual-fairness.html +++ b/explainers/counterfactual-fairness.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Counterfactual Fairness
Would the outcome change if identity changed?
diff --git a/explainers/cross-validation-for-fairness.html b/explainers/cross-validation-for-fairness.html index ea296081..d56b3768 100644 --- a/explainers/cross-validation-for-fairness.html +++ b/explainers/cross-validation-for-fairness.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Cross-Validation for Fairness
Test every row once, and protect the small cells.
diff --git a/explainers/data-leakage.html b/explainers/data-leakage.html index e318a029..3e5fa3f9 100644 --- a/explainers/data-leakage.html +++ b/explainers/data-leakage.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
What Is Data Leakage?
When the model has already seen the answer sheet.
diff --git a/explainers/datasheets-for-datasets.html b/explainers/datasheets-for-datasets.html index 58d7effc..2ac9c1b9 100644 --- a/explainers/datasheets-for-datasets.html +++ b/explainers/datasheets-for-datasets.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Datasheets for Datasets
Ask the questions about a dataset that nobody wrote down.
diff --git a/explainers/decision-curve-analysis-by-group.html b/explainers/decision-curve-analysis-by-group.html index 6509d03a..79f9ee1c 100644 --- a/explainers/decision-curve-analysis-by-group.html +++ b/explainers/decision-curve-analysis-by-group.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
What Is Decision Curve Analysis by Group?
The thresholds at which a model helps one group can be the ones where it fails another.
diff --git a/explainers/delayed-feedback.html b/explainers/delayed-feedback.html index 252d99cc..089b2d34 100644 --- a/explainers/delayed-feedback.html +++ b/explainers/delayed-feedback.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Delayed Feedback
Outcomes that have not happened yet look like outcomes that did not happen.
diff --git a/explainers/demographic-parity.html b/explainers/demographic-parity.html index 2a491083..1dcb7712 100644 --- a/explainers/demographic-parity.html +++ b/explainers/demographic-parity.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Demographic Parity
Equal positive rates across groups.
diff --git a/explainers/differential-privacy.html b/explainers/differential-privacy.html index 1d9c118e..1b2885c2 100644 --- a/explainers/differential-privacy.html +++ b/explainers/differential-privacy.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Differential Privacy and Its Tension With Fairness
The accuracy cost of a privacy guarantee falls hardest on underrepresented groups.
diff --git a/explainers/differential-validity.html b/explainers/differential-validity.html index 65b08163..1bce2e8f 100644 --- a/explainers/differential-validity.html +++ b/explainers/differential-validity.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Differential Validity
A score can predict equally well and still be set too high for one group.
diff --git a/explainers/disaggregated-evaluation.html b/explainers/disaggregated-evaluation.html index 176790b5..219a9667 100644 --- a/explainers/disaggregated-evaluation.html +++ b/explainers/disaggregated-evaluation.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Disaggregated Evaluation
One overall number hides who the model fails.
diff --git a/explainers/discharge-leakage-in-readmission-models.html b/explainers/discharge-leakage-in-readmission-models.html index 74e7500d..e6ed9655 100644 --- a/explainers/discharge-leakage-in-readmission-models.html +++ b/explainers/discharge-leakage-in-readmission-models.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Discharge Leakage in Readmission Models
Some columns are only known once the outcome is nearly decided.
diff --git a/explainers/disparate-impact.html b/explainers/disparate-impact.html index 2d00f199..abdaeafa 100644 --- a/explainers/disparate-impact.html +++ b/explainers/disparate-impact.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Disparate Impact
When outcomes differ even without explicit intent.
diff --git a/explainers/disparate-treatment.html b/explainers/disparate-treatment.html index ce8884e7..80c9301a 100644 --- a/explainers/disparate-treatment.html +++ b/explainers/disparate-treatment.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Disparate Treatment
Direct discrimination instead of proxy discrimination.
diff --git a/explainers/distribution-shift.html b/explainers/distribution-shift.html index 6394ee5c..d966e622 100644 --- a/explainers/distribution-shift.html +++ b/explainers/distribution-shift.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Distribution Shift
A fairness audit is only valid for the data it was run on
diff --git a/explainers/distributionally-robust-optimization.html b/explainers/distributionally-robust-optimization.html index 83d3fb94..14f5c17b 100644 --- a/explainers/distributionally-robust-optimization.html +++ b/explainers/distributionally-robust-optimization.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Distributionally Robust Optimization for Fairness
Hedge against the worst group instead of constraining one metric.
diff --git a/explainers/equal-opportunity.html b/explainers/equal-opportunity.html index 798925f5..6009da68 100644 --- a/explainers/equal-opportunity.html +++ b/explainers/equal-opportunity.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Equal Opportunity vs Equalized Odds
Passing the true-positive-rate check doesn't mean passing the false-positive-rate one.
diff --git a/explainers/equalized-odds.html b/explainers/equalized-odds.html index 5f0a1974..3687d7f9 100644 --- a/explainers/equalized-odds.html +++ b/explainers/equalized-odds.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Equalized Odds
A fairness metric that checks error rates, not just accuracy.
diff --git a/explainers/explanation-disagreement.html b/explainers/explanation-disagreement.html index a39e68f5..ab6b5ff4 100644 --- a/explainers/explanation-disagreement.html +++ b/explainers/explanation-disagreement.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Explanation Disagreement
Two explanation methods, one model, different top features.
diff --git a/explainers/fair-representation-learning.html b/explainers/fair-representation-learning.html index 51a0681e..5cf655fb 100644 --- a/explainers/fair-representation-learning.html +++ b/explainers/fair-representation-learning.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Fair Representation Learning
A representation that hides the group from one reader can still leak to another.
diff --git a/explainers/fairness-accuracy-pareto-frontier.html b/explainers/fairness-accuracy-pareto-frontier.html index a977559a..19db8b6f 100644 --- a/explainers/fairness-accuracy-pareto-frontier.html +++ b/explainers/fairness-accuracy-pareto-frontier.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Fairness-Accuracy Pareto Frontier
The set of choices nothing else beats, and it can be short.
diff --git a/explainers/fairness-accuracy-tradeoff.html b/explainers/fairness-accuracy-tradeoff.html index fa84f3ca..879f1ab9 100644 --- a/explainers/fairness-accuracy-tradeoff.html +++ b/explainers/fairness-accuracy-tradeoff.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
What Is the Fairness-Accuracy Trade-off?
Closing a fairness gap is rarely free, and the cost isn't always the same size.
diff --git a/explainers/fairness-in-ranking.html b/explainers/fairness-in-ranking.html index 8c554c21..3039c58e 100644 --- a/explainers/fairness-in-ranking.html +++ b/explainers/fairness-in-ranking.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Fairness in Ranking
Two similar scores can still mean very different attention.
diff --git a/explainers/fairness-metric-conflicts.html b/explainers/fairness-metric-conflicts.html index 130e76bb..34430887 100644 --- a/explainers/fairness-metric-conflicts.html +++ b/explainers/fairness-metric-conflicts.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Fairness Metric Conflicts
Why one fairness definition often breaks another.
diff --git a/explainers/fairness-regularization.html b/explainers/fairness-regularization.html index 1d49c15d..f3127329 100644 --- a/explainers/fairness-regularization.html +++ b/explainers/fairness-regularization.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Fairness Regularization
Put a price on the parity gap and read the trade-off off a curve.
diff --git a/explainers/fairness-through-unawareness.html b/explainers/fairness-through-unawareness.html index c6b0f61d..b16ae02e 100644 --- a/explainers/fairness-through-unawareness.html +++ b/explainers/fairness-through-unawareness.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Why Fairness Through Unawareness Fails
Removing the protected attribute doesn't remove what correlates with it.
diff --git a/explainers/fairness-without-demographics.html b/explainers/fairness-without-demographics.html index a9cd89bb..6c7acede 100644 --- a/explainers/fairness-without-demographics.html +++ b/explainers/fairness-without-demographics.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Fairness Without Demographics
Help the worst-off group when you do not know who they are.
diff --git a/explainers/false-positives-vs-false-negatives.html b/explainers/false-positives-vs-false-negatives.html index cbb5d241..03c6b636 100644 --- a/explainers/false-positives-vs-false-negatives.html +++ b/explainers/false-positives-vs-false-negatives.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
False Positives vs. False Negatives in Medical AI
When a missed high-risk flag costs more than a false alarm
diff --git a/explainers/feature-importance-by-group.html b/explainers/feature-importance-by-group.html index 64d40df2..55f65cfb 100644 --- a/explainers/feature-importance-by-group.html +++ b/explainers/feature-importance-by-group.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Feature Importance by Group
Does the model lean on different features for different people?
diff --git a/explainers/feedback-loop-bias.html b/explainers/feedback-loop-bias.html index c42c4257..4cb33ed3 100644 --- a/explainers/feedback-loop-bias.html +++ b/explainers/feedback-loop-bias.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Feedback Loop Bias
Bias that gets stronger after deployment.
diff --git a/explainers/geographic-proxies-and-redlining.html b/explainers/geographic-proxies-and-redlining.html index e0e9e3f0..e1eb4412 100644 --- a/explainers/geographic-proxies-and-redlining.html +++ b/explainers/geographic-proxies-and-redlining.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Geographic Proxies and Redlining
Where a person lives is the oldest proxy for who they are.
diff --git a/explainers/group-specific-thresholds.html b/explainers/group-specific-thresholds.html index 4b81b494..7a85f861 100644 --- a/explainers/group-specific-thresholds.html +++ b/explainers/group-specific-thresholds.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Group-Specific Thresholds
Keep the model, move the cutoff.
diff --git a/explainers/hiring-pipeline-adverse-impact.html b/explainers/hiring-pipeline-adverse-impact.html index 7292c424..ddf69e75 100644 --- a/explainers/hiring-pipeline-adverse-impact.html +++ b/explainers/hiring-pipeline-adverse-impact.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Hiring Pipeline Adverse Impact
Adverse impact at each stage, not only the last decision.
diff --git a/explainers/how-ai-detects-patterns.html b/explainers/how-ai-detects-patterns.html index 4f856ab5..e5d5c39a 100644 --- a/explainers/how-ai-detects-patterns.html +++ b/explainers/how-ai-detects-patterns.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
How AI Detects Patterns
A model can't tell a cause from a proxy
diff --git a/explainers/individual-fairness.html b/explainers/individual-fairness.html index d4013125..3eca3e06 100644 --- a/explainers/individual-fairness.html +++ b/explainers/individual-fairness.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Individual Fairness
Similar people should get similar outcomes.
diff --git a/explainers/inferring-protected-attributes.html b/explainers/inferring-protected-attributes.html index 036f803f..4d521ce5 100644 --- a/explainers/inferring-protected-attributes.html +++ b/explainers/inferring-protected-attributes.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Inferring Protected Attributes
Measuring a gap when the group is not in the data.
diff --git a/explainers/intersectional-bias.html b/explainers/intersectional-bias.html index b5989728..18e8a2e7 100644 --- a/explainers/intersectional-bias.html +++ b/explainers/intersectional-bias.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
What Is Intersectional Bias?
Checking one protected attribute at a time can hide harm concentrated at the intersection.
diff --git a/explainers/label-bias.html b/explainers/label-bias.html index 78b7d34d..8bd65f19 100644 --- a/explainers/label-bias.html +++ b/explainers/label-bias.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Label Bias
When the target itself is already skewed.
diff --git a/explainers/label-shift-and-fairness.html b/explainers/label-shift-and-fairness.html index c77ad3cb..233313f7 100644 --- a/explainers/label-shift-and-fairness.html +++ b/explainers/label-shift-and-fairness.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Label Shift and Fairness
When the base rate moves, some metrics swing and some stay put.
diff --git a/explainers/leave-one-group-out-evaluation.html b/explainers/leave-one-group-out-evaluation.html index f861a851..3306da16 100644 --- a/explainers/leave-one-group-out-evaluation.html +++ b/explainers/leave-one-group-out-evaluation.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Leave-One-Group-Out Evaluation
How does the model do on a group it never saw?
diff --git a/explainers/lime.html b/explainers/lime.html index 8c8274a6..1624def1 100644 --- a/explainers/lime.html +++ b/explainers/lime.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
What Is LIME?
A local, approximate surrogate model - the other major way to explain one prediction.
diff --git a/explainers/maxmin-fairness.html b/explainers/maxmin-fairness.html index d40f2092..56763fc4 100644 --- a/explainers/maxmin-fairness.html +++ b/explainers/maxmin-fairness.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Max-Min (Rawlsian) Fairness
Make the worst-off group as well off as possible, even if groups end up unequal.
diff --git a/explainers/measurement-bias.html b/explainers/measurement-bias.html index 5248480d..a94ec899 100644 --- a/explainers/measurement-bias.html +++ b/explainers/measurement-bias.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Measurement Bias
When the feature or label measures one group worse than another.
diff --git a/explainers/medical-imaging-representation-gaps.html b/explainers/medical-imaging-representation-gaps.html index 5a1abb0f..3a74741f 100644 --- a/explainers/medical-imaging-representation-gaps.html +++ b/explainers/medical-imaging-representation-gaps.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Why Medical Imaging Models Fail Some Groups
A model trained mostly on one group's images has barely seen the others.
diff --git a/explainers/minimum-detectable-gap.html b/explainers/minimum-detectable-gap.html index acdaa6a4..6d95a6b5 100644 --- a/explainers/minimum-detectable-gap.html +++ b/explainers/minimum-detectable-gap.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Minimum Detectable Gap
A non-significant gap can be a gap the audit could never see.
diff --git a/explainers/missing-data-bias-ehr.html b/explainers/missing-data-bias-ehr.html index 464fbc6f..db1aad1b 100644 --- a/explainers/missing-data-bias-ehr.html +++ b/explainers/missing-data-bias-ehr.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Missing Data as Bias in Electronic Health Records
A patient with fewer recorded labs and visits is less-observed, not lower-risk.
diff --git a/explainers/mitigation-strategies.html b/explainers/mitigation-strategies.html index 2d2f95f1..1810e86f 100644 --- a/explainers/mitigation-strategies.html +++ b/explainers/mitigation-strategies.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Pre-, In-, and Post-Processing Fairness Mitigations
Where in the pipeline you intervene changes what a fairness fix can and can't do.
diff --git a/explainers/ml-bias.html b/explainers/ml-bias.html index 58ab1375..98a62e7d 100644 --- a/explainers/ml-bias.html +++ b/explainers/ml-bias.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
What Is Machine Learning Bias?
Four entry points. One pipeline. Measurable everywhere.
diff --git a/explainers/model-cards.html b/explainers/model-cards.html index 56715cd1..08fceacc 100644 --- a/explainers/model-cards.html +++ b/explainers/model-cards.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Model Cards
A short, standard disclosure of what a model is for and who it fails.
diff --git a/explainers/model-drift.html b/explainers/model-drift.html index 69989619..dde3995d 100644 --- a/explainers/model-drift.html +++ b/explainers/model-drift.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
What Is Model Drift?
A fairness audit is a photograph of a moving room
diff --git a/explainers/monitoring-fairness-in-production.html b/explainers/monitoring-fairness-in-production.html index cdebb538..6686999f 100644 --- a/explainers/monitoring-fairness-in-production.html +++ b/explainers/monitoring-fairness-in-production.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Monitoring Fairness in Production
An audit is a photograph. Monitoring is the film.
diff --git a/explainers/multiple-comparisons.html b/explainers/multiple-comparisons.html index 7428644f..bb7c1601 100644 --- a/explainers/multiple-comparisons.html +++ b/explainers/multiple-comparisons.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Multiple-Comparisons Correction for Fairness Audits
Run enough p-values at 0.05 and some will read "significant" by chance alone.
diff --git a/explainers/neural-networks.html b/explainers/neural-networks.html index a2764190..325d8876 100644 --- a/explainers/neural-networks.html +++ b/explainers/neural-networks.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Neural Networks
How complex models can hide simple bias.
diff --git a/explainers/obermeyer-cost-proxy.html b/explainers/obermeyer-cost-proxy.html index 687b48c7..e698ec4d 100644 --- a/explainers/obermeyer-cost-proxy.html +++ b/explainers/obermeyer-cost-proxy.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
The Obermeyer Case: Cost as a Proxy for Health Need
How predicting healthcare spending instead of illness systematically under-refers sicker Black patients.
diff --git a/explainers/partial-dependence-by-group.html b/explainers/partial-dependence-by-group.html index ad0e1f6e..576f80bf 100644 --- a/explainers/partial-dependence-by-group.html +++ b/explainers/partial-dependence-by-group.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Partial Dependence by Group
What the model does with a feature, drawn once per group.
diff --git a/explainers/path-specific-effects.html b/explainers/path-specific-effects.html index b18ea831..9da6b135 100644 --- a/explainers/path-specific-effects.html +++ b/explainers/path-specific-effects.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Path-Specific Effects
Split a gap into the paths you accept and the ones you do not.
diff --git a/explainers/permutation-test-fairness-gaps.html b/explainers/permutation-test-fairness-gaps.html index de1b373f..560049e2 100644 --- a/explainers/permutation-test-fairness-gaps.html +++ b/explainers/permutation-test-fairness-gaps.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
Permutation Tests for Fairness Gaps
Is the gap a finding or a coincidence?
diff --git a/explainers/popularity-bias.html b/explainers/popularity-bias.html index 18fedef8..4f1edc29 100644 --- a/explainers/popularity-bias.html +++ b/explainers/popularity-bias.html @@ -173,6 +173,7 @@ + Skip to contentExplainer
What Is Popularity Bias?
Items that were shown get the clicks that keep them shown; the rest never get a chance.
diff --git a/explainers/position-bias-and-click-models.html b/explainers/position-bias-and-click-models.html index 6bc84fe8..2bb5328b 100644 --- a/explainers/position-bias-and-click-models.html +++ b/explainers/position-bias-and-click-models.html @@ -173,6 +173,7 @@ + Skip to content