Bias in AI

This repository contains a functional AIMMS example model for Bias in AI. It demonstrates how to connect AIMMS to a Python machine learning service and expose potential bias in AI-driven toxicity classification.

🎯 Business Problem

The combination of machine learning and everyday applications is at the heart of modern tech advancements. But hidden beneath its brilliance is a complex issue — bias within these algorithms.

This example illustrates bias by creating an AIMMS front-end to an existing Python application based on Kaggle's AI Ethics course. The application:

• Accepts a user-entered comment and evaluates its toxicity.
• Reads in a training dataset of comments with toxicity labels.
• Passes both the training data and the new comment to a Python service.
• Returns whether the comment is considered toxic — revealing that the underlying model may treat identical concepts differently depending on the demographic group referenced.

Note: Bias can be observed in practice by entering words like black (marked toxic) versus white (marked not toxic).

This is also a relevant concern for Decision Support applications: basing decisions on data that is not representative of your market leads to poor outcomes.

📖 How to Use This Example

To get the most out of this model, we highly recommend reading our detailed step-by-step guide on the AIMMS How-To website:

👉 Read the Full Article: Bias in AI

Prerequisites

• AIMMS: You will need AIMMS installed to run the model. An AIMMS Community License is sufficient. Download the Free Academic Edition here if you are a student.
• Python 3.11: Required to run the Python toxicity classification service. PyCharm is recommended but not required.
• WebUI: This model is optimized for the AIMMS WebUI for a modern, browser-based experience.

🚀 Getting Started

1. Download the Release: Go to the Releases page and download the .zip file from the latest version.
2. Open the Project: Launch the .aimms file.
3. Start the Python Service: Follow the article instructions to start the Python backend locally or on AIMMS Cloud.
4. Explore Bias: Use the WebUI to import the dataset, enter comments, and observe the toxicity classification results.

🤝 Support & Feedback

This example is maintained by the AIMMS User Support Team.

• Found an issue? Open an issue.
• Questions? Reach out via the AIMMS Community.

---

Maintained by the AIMMS User Support Team. We optimize the way you build optimization.