Module 5: Algorithmic Ethics

---

Explainability and Transparency

Trust and transparency are often linked, for good reason. However, when it comes to AI, the linkage is often assumed that if the AI can somehow explain its reasoning, people will automatically have more trust in it. Creating trustworthy explanations of AI is more challenging than you might think! For this assignment, we will again discuss a small set of papers focused on transparency and explainability.

Assignment 1

For this assignment, I have a few papers to already choose from or you can again look at the FAACT conference.

• You can read “How do I fool you?”: Manipulating User Trust via Misleading Black Box Explanations” 
• You can read “The Road to Explainability is Paved with Bias: Measuring the Fairness of Explanations“
• The two conferences below are where the papers appeared so you are also welcome to look at them!
  • Look at the past few years of the “ACM Conference on Fairness, Accountability, and Transparency” and choose a paper that is of interest to you focused on explainability/transparency. Read it and prepare a summary for class.
  • Look at the past few years of the “AAAI/ACM Conference on AI, Ethics, and Society” and find a paper on explainability/transparency that looks interesting to you. Read and prepare a summary for class.