← All projects

PROJECT 02 / Responsible AI · Explainability

Loan Bias Auditor

A model-auditing tool that examines differences in loan-approval outcomes and makes fairness findings easier to inspect through visual reports.

01

The problem

A loan-approval model can produce different outcomes across demographic groups. Aggregate performance alone does not explain whether those differences warrant closer review.

02

My approach

Built an auditing system that compares loan-approval outcomes across groups, including dimensions such as gender, age, income and employment status.

03

How it works

Implemented approval-rate difference, demographic parity, equal opportunity, feature importance and SHAP analysis, supported by visual reports and bias alerts.

04

What it demonstrates

Using fairness measures and model explanations to surface patterns that deserve investigation, and communicating those patterns through visual reports.

05

Limitations

Fairness measures answer different questions and can disagree. Findings depend on the data, group definitions and decision thresholds. An alert does not by itself establish discrimination or certify that a model is fair.

KEEP EXPLORING

TruthLens AIStock Market Analysis & PredictionSri Balaji Earth Movers

Have a similar question?

Let’s talk