Ethical AI Design: Principles and Implementation

This module provides an overview of ethical challenges in AI and introduces the Design for Values methodology. Students will learn to identify values of different stakeholders and methods to operationalize these, applied to a central case study in healthcare.

This module focuses on the trustworthiness of AI systems, exploring concepts of accuracy and explainability. Students will examine when AI systems are considered trustworthy and how this relates to requirements for accuracy and explainability. The module also covers available technical tools for improving system explainability.

This module addresses bias in data and algorithmic fairness. Students will explore philosophical conceptions of fairness and bias, as well as concrete metrics and tools for monitoring and correcting biases. The module also covers the identification of problematic statistical biases and appropriate steps to address them.

This module examines accountability and human oversight in AI systems. Students will explore questions of responsibility when mistakes occur with AI systems and discuss the organizational and socio-technical design needed for responsible AI use. The concept of meaningful human control is introduced, along with various implementation methods for oversight.

This module explores value conflicts in AI ethics. Students will learn how to navigate situations where different ethical values are challenging to realize simultaneously. The module includes a final assignment where students apply the process of translating ethical values into design requirements for a new case study.