Interactive Explainable AI
summary
Argues that interactive XAI uncovers what models actually use and reduces biases by letting users manipulate inputs and see outputs. Demonstrates feature-attribution and example-based methods: What-If Tool sliders change predictions live, the Language Interpretability Tool highlights influential words, and a Smiley nearest-neighbor explorer lets pathologists focus on regions. Reports a concrete result: remapping radar colors to align with perceptual jumps at 36 DBZ threshold and 55 DBZ raises perceived forecast accuracy.
AI-written summary · about
chapters (7)
- 0:00Intro
- 1:59What do we need XAI for?
- 4:33Human factors of AI and XAI
- 9:36We’ve actually been explaining complex things for a long time
- 15:43The UX of XAI
- 23:07Examples of interactive XAI
- 27:18Parting Thoughts
related talks
39mThe Search for Transparency and Accountability in the Age of...
46mCan you trust your AI?
32mOn the Importance of Explainability
39mKeynote 3: How did you know? Explaining Black Box Model Predictions in Python
45mAI-404: Explanation Not Found
10mI Think I Get Your Point, AI! The Illusion of Explanatory Depth in Explainable AI
19mMoving Forward with Explainable AI: Adoption Rate & Essential Dimensions for a Reliable Use
56mAI: It AI-n't What You Think!