Fairness Metrics Computation Tool for AI Model Bias Analysis

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Measuring and quantifying algorithmic bias in AI/ML models is non-trivial due to their complex, non-linear nature and the numerous input variables, making existing software tools require many manual data processing steps and lack standardization.

Innovation Solution

A software tool with a graphical user interface (GUI) that allows users to input data from AI/ML models, identify prediction, ground truth, and sensitive attribute columns, and automatically generate algorithmic fairness analysis and visual reports, reducing manual effort and increasing standardization.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If manual data processing steps are used in existing software tools, then flexibility in handling diverse data can be maintained, but the complexity and time required for analysis increases significantly

Engineering Contradiction:

Improveease of operationVSAvoidtime required for analysis

Core Design Contradiction:

Ease of operationVSLoss of time

Solution Approach 1:

The system performs preliminary actions by automatically detecting data types, identifying columns, and preparing data structures before the actual fairness analysis. The GUI automatically recognizes prediction columns, ground truth columns, and sensitive attribute columns without requiring manual preprocessing, thus reducing the time and effort needed for analysis while maintaining operational ease

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system enables self-service by automatically processing and analyzing the uploaded data without requiring extensive manual intervention. The fairness analysis tool autonomously computes metrics, generates reports, and provides visualizations based on the input data, eliminating the need for manual data processing steps while maintaining flexibility in handling diverse data formats

Inventive Principle:

Principle #25Self-service

2Reliability

If standardized analysis processes are implemented, then reproducibility and comparability improve, but adaptability to different AI/ML models decreases

Engineering Contradiction:

ImprovereproducibilityVSAvoidadaptability to different models

Core Design Contradiction:

ReliabilityVSAdaptability or versatility

Solution Approach 1:

The system achieves universality by designing a unified fairness analysis framework that can handle multiple types of AI/ML models through a single standardized interface. The GUI accepts various data formats and model outputs, automatically adapts to different column structures, and applies consistent fairness metrics across diverse models, thus maintaining both reproducibility and adaptability simultaneously

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The system employs parameter changes by allowing users to configure analysis parameters such as selection of sensitive attributes, choice of fairness metrics, and data processing options. This flexibility in parameter configuration enables the standardized process to adapt to different AI/ML models while maintaining consistent and reproducible analysis procedures

Inventive Principle:

Principle #35Parameter changes

3Measurement precision

If comprehensive fairness metrics are computed, then the accuracy of bias measurement improves, but the computational complexity and resource requirements increase

Engineering Contradiction:

Improveaccuracy of bias measurementVSAvoidcomputational complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The system applies segmentation by dividing the comprehensive fairness analysis into distinct modular components. The GUI separates the computation of different fairness metrics (e.g., demographic parity, equalized odds, calibration) into independent modules that can be executed sequentially or in parallel, reducing overall computational complexity while maintaining measurement precision through comprehensive metric computation

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS20250077939A1Systems and methods for computing fairness metrics in artificial intelligence and machine learning models

Publication Date: 2025.03.06 THE TORONTO DOMINION BANK

AI summary

Systems and methods for assessing machine learning model fairness are provided. A graphical user interface (GUI) is rendered in a web browser. Input data from the web browser is obtained. A first input produced by a first GUI control is received to identify a prediction column in the input data. A second input produced by a second GUI control is received to identify a ground truth column. A third input produced by a third GUI control is received to identify a sensitive attribute column. A plurality of groups within a plurality of sensitive attribute values in the sensitive attribute column are automatically displayed. The system processes the input data based on the prediction column, the ground truth column and the sensitive attribute column to generate one or more algorithmic fairness analysis, which are used to display a visual report.