Book Image

Responsible AI in the Enterprise

By : Adnan Masood, Heather Dawe
5 (1)
Book Image

Responsible AI in the Enterprise

5 (1)
By: Adnan Masood, Heather Dawe

Overview of this book

Responsible AI in the Enterprise is a comprehensive guide to implementing ethical, transparent, and compliant AI systems in an organization. With a focus on understanding key concepts of machine learning models, this book equips you with techniques and algorithms to tackle complex issues such as bias, fairness, and model governance. Throughout the book, you’ll gain an understanding of FairLearn and InterpretML, along with Google What-If Tool, ML Fairness Gym, IBM AI 360 Fairness tool, and Aequitas. You’ll uncover various aspects of responsible AI, including model interpretability, monitoring and management of model drift, and compliance recommendations. You’ll gain practical insights into using AI governance tools to ensure fairness, bias mitigation, explainability, privacy compliance, and privacy in an enterprise setting. Additionally, you’ll explore interpretability toolkits and fairness measures offered by major cloud AI providers like IBM, Amazon, Google, and Microsoft, while discovering how to use FairLearn for fairness assessment and bias mitigation. You’ll also learn to build explainable models using global and local feature summary, local surrogate model, Shapley values, anchors, and counterfactual explanations. By the end of this book, you’ll be well-equipped with tools and techniques to create transparent and accountable machine learning models.
Table of Contents (16 chapters)
Part 1: Bigot in the Machine – A Primer
Part 2: Enterprise Risk Observability Model Governance
Part 3: Explainable AI in Action

Bias and disparity mitigation with Fairlearn

Fairlearn provides several ways to perform bias and disparity mitigation for real-world problems:

  • Post-processing methods: This involves adjusting the predictions made by a machine learning model after it has been trained, to reduce bias and disparity. An example of this is the reject option classifier, which allows you to set a threshold for the prediction scores for certain sensitive features. If the threshold is exceeded, the classifier will reject the prediction and instead return a default label.
  • Pre-processing methods: This involves transforming the data before training the machine learning model, to reduce bias and disparity. An example of this is CorrelationRemover, which adjusts the non-sensitive features to remove their correlation with the sensitive features, while retaining as much information as possible.
  • In-processing methods: This involves modifying the training process of the machine learning model, to reduce...