About EthicAI

EthicAI is an educational fairness monitoring tool that evaluates bias in model predictions across protected attributes (e.g., sex, race). It is designed to support educators, researchers, and practitioners with clear, visual summaries for non-technical audiences.

Ethical Framework

Use only public, de-identified data and respect dataset licensing rules.
Acknowledge representation imbalance and historical bias in data.
Frame results as indicators for investigation, not proof of discrimination.
Explain metrics in accessible language for broad audiences.

Fairness Metric Definitions

Selection Rate: The fraction of predictions labeled “positive”.

TPR (True Positive Rate): Of truly positive cases, the fraction predicted positive.

FPR (False Positive Rate): Of truly negative cases, the fraction predicted positive.

FNR (False Negative Rate): Of truly positive cases, the fraction predicted negative.

Demographic Parity Difference (DPD): The max gap in selection rates between groups.

Equal Opportunity Difference (EOD): The max gap in TPR between groups.

This tool does not make hiring, salary, or legal decisions. It supports transparency and responsible evaluation.