Core

Intersectional and Subgroup Fairness

A model can pass a fairness audit on gender, pass on race, and fail badly on their intersection, because a single-axis audit averages away the group you most need to see. Doing it properly means fighting combinatorial explosion, small noisy cells, multiple-comparison error, and fairness gerrymandering (fair on every named group, unfair on one nobody named). AI, ML, and GenAI interviews probe it because reporting per-group metrics one axis at a time is the standard answer and it is not sufficient.

a free account unlocks the core curriculum tier · no card

RELATED CONCEPTS

Fairness, Bias, and Model Cards Calibration and Uncertainty Precision, Recall, and F1 AI Governance Frameworks

PRACTICE THIS IN REAL QUESTIONS

AI Security, Privacy & GovernanceHow do you detect and mitigate bias in an ML model used for consequential decisions?→AI Security, Privacy & GovernanceYour model denies someone a loan, and they demand to know why. How do you handle the right to explanation?→AI Security, Privacy & GovernanceThreat-model an LLM application from scratch. What is the attack surface and how do you reason about it?→AI Security, Privacy & GovernanceYour model passes bias checks for gender and for race separately, but fails for Black women. How do you handle intersectional fairness?→AI Security, Privacy & GovernanceHow do you handle PII and data governance for an enterprise LLM deployment (SOC 2, GDPR, the EU AI Act)?→RAG & Agent System DesignDesign multi-tenancy and access control for a RAG system serving many enterprise customers.→

COMPANIES THAT ASSUME THIS

Anthropic IBM SSI (Safe Superintelligence)