Bias and Discrimination in AI

The use of Machine Learning, and Artificial Intelligence in general, has been extensively applied to many sectors to enhance decision making and early detection. In Healthcare, the use of big data and automated decision-making extends to hospital management, disease detection and health prevention. Nevertheless, the use of artificial intelligence carries issues that can affect individuals’ life and limits the access to healthcare and particular treatments.

In line with the mission of the MSc in Health Management & Data Intelligence program, this couse integrates work that deals with bias and discrimination in Artificial Intelligence philosophically and practically. This course is oriented towards a critical approach to AI and analyses the main factors that increase the risk of social discrimination. The course covers a legal approach to EU regulation, a technical approach that covers techniques for detecting and mitigating biases, and also, provides an overview of current debates in the field of artificial intelligence.

The goal of this course is to provide tools and main conceptual frameworks to discuss and implement strategies for prevent and mitigate discrimination of AI and Big Data. It is expected that students:

• Understand the main concepts behind European regulation on algorithmic discrimination

• Understand and make use of fairness metrics in the context of the healthcare sector

• Develop a critical mindset on how data and algorithms can affect people and social groups

Contents

Module 1: Privacy, ethics and regulation on discrimination

Module 2: Bias and types of harms in data

Module 3: Algorithmic fairness in classification problems

Module 4: Algorithmic fairness in other fields of Machine Learning

Module 5: Intersectional analysis in AI

Module 6: Auditing, impact assessments and Human Oversight