After completing this course you should be able to:

• Identify and anticipate the types of unintended harm that can arise from AI models
• Explain why interpretability is key to avoiding harm
• Distinguish between intrinsically interpretable models and black box models
• Evaluate tradeoffs between model performance and interpretability
• Establish a Responsible Data Science framework for your projects

This course is arranged in 4 modules. We estimate that you will need 5 hours per week. The course is self-paced, so you have the flexibility to complete the modules in your own time.

Week 1 – Landscape of Harm

• Videos:

  • AI and Big Brother

  • Unintended harm

  • Types of harm

  • Best Practices - CRISP-DM

  • A bit of ancient history (verified users only)

• Knowledge Checks

• Reading / Discussion Prompt 1

• Exercise 1 & 2 (for verified users only)

Week 2 – Legal Issues

• Videos:

  • Legal Issues EU

  • Existing laws

• Knowledge Checks

• Reading / Discussion Prompt 2

• Exercise 3 (for verified users only)

Week 3 – Transparency

• Videos:

  • Model interpretability

  • Global interpretability methods

• Knowledge Checks

• Reading

• Exercise 4 & 5 (for verified users only)

Week 4 – Principles and Frameworks

• Videos:

  • Introduction to Principles of Responsible Data Science (RDS)

  • From Principles to Practice

  • RDS Framework

  • Return to CRISP-DM

• Knowledge Checks

• Exercise 6 (for verified users only)