Robustness, Transparency, and Security

This unit examines how to design and assess trustworthy AI through robustness, transparency, explainability, and security. Students build on foundational AI concepts to weigh trade-offs, apply explainability techniques, and identify security compromises, preparing them to lead governance and mitigation efforts in subsequent units.

Learning Objectives

• Analyze the limits and benefits of transparency in AI systems and evaluate trade-offs between interpretability, performance, and stakeholder needs
• Evaluate stakeholder impacts by applying ethical and communication criteria to determine when and how to disclose model behavior and decision rationale
• Apply technical and process-oriented approaches to increase explainability and reliability, including selecting appropriate explainability methods and documenting design decisions
• Demonstrate the ability to identify common AI compromise scenarios and propose security controls and mitigation strategies to reduce exploitability