AI Fairness & Bias Metrics: Parity, Equal Opportunity & Calibration by Group

About this course

An AI fairness course that audits one fictional loan-approval model group by group in a spreadsheet, for data scientists, ML engineers, risk and compliance teams and students of responsible AI. You count true and false positives and negatives for each group with COUNTIFS and compare accuracy and base rates, calculate selection rates and the demographic parity difference, compute the disparate impact ratio and flag groups below the four-fifths threshold, measure true and false positive rate gaps for equal opportunity and equalised odds, check calibration and predictive parity by score bin, compare four threshold policies on parity and accuracy, split groups by age band with margins of error and flag subgroups too small to judge, and finish with a flagged fairness report card. The four-fifths rule is used as a screening heuristic, not a legal test, and the course is educational, not legal advice. The first lab is free.