Home

Fujitsu AI Ethics for Fairness

We aim to realize a fair and trustworthy AI society through technologies that ensure the fairness of AI.

In ensuring the fairness of AI, biases related to race and gender pose significant problems. Detecting and mitigating these biases requires specialized knowledge, which places a heavy burden on AI developers.

With Fujitsu AI Ethics for Fairness

Essential functions can be executed easily through a web browser, enabling even developers without expertise in AI ethics to verify fairness.
Thanks to our unique intersectional bias mitigation technology, the system can detect and mitigate subtle biases that arise from combinations of multiple attributes—biases that are often difficult for humans to notice.

The Value Brought by Fujitsu AI Ethics for Fairness

Fairness Verification Made Intuitive

From data upload and fairness verification to bias mitigation, model training, and performance comparison—everything is seamlessly achievable.

The system can detect and mitigate intersectional biases that emerge when attributes such as age, gender, or nationality are combined under specific conditions.

Detection and Mitigation of Subtle Biases

The system can detect and mitigate intersectional biases that emerge when attributes such as age, gender, or nationality are combined under specific conditions.

Demo
App Click here to try the demo app

Technical Overview

Target Industries / Users

Companies developing AI systems—especially those handling human-related data, such as in loan screening or recruitment.

Challenges in Target Industries and Operations

In developing AI systems, ensuring the fairness of AI is a requirement.

Technical Challenges

It should be possible for non-experts in AI ethics and those without programming skills to verify and improve AI fairness.
There are biases that are difficult for humans to detect, which need to be identified and addressed.

Solution

Provide a user interface that allows intuitive operation via a web browser, enabling users to easily perform a series of tasks such as data upload, fairness verification, and AI model comparison.
Automatically detect and mitigate intersectional biases that arise from combinations of multiple attributes and are often difficult for humans to notice.

Fujitsu's Technological Advantage

Fujitsu's unique bias mitigation technology enables the effortless detection and improvement of intersectional bias, addressing trends that vary across combinations of multiple attributes (e.g., gender × age).

Detailed Value of Fujitsu AI Ethics for Fairness

In AI system development, even developers without advanced knowledge of AI ethics can ensure fairness by using this technology.

Usage scenario

During AI system development, developers can verify the fairness of the AI using the tool.

Technical Trial

Demo App：The demo app is available for you to use.
Demo video
A Proof of Concept (PoC) is possible.

Press Release
- September 15, 2023 "Fujitsu and the Linux Foundation launch Fujitsu's automated machine learning and AI fairness technologies as Linux Foundation hosted open source projects"

Documents

Document Title	Explanation
Sample data	Sample data to use when trying out Fujitsu AI Ethics for Fairness.
License information	License information for OSS, etc. used by Fujitsu AI Ethics for Fairness.