Explaining the Explainable AI: A 2-Stage Approach - Dr. Amit Dhurandhar
Explainable AI (or XAI) has garnered a lot of interest in recent years across academia, industry and government. Although many methods have been proposed it is still unclear what XAI truly entails and why it is hard to formalize as opposed to other areas of machine learning such as causality, adversarial robustness, amongst others. In this talk, I will try to explicitly point out what XAI is trying to do, thus making it clear why formalization is difficult. The disentangled perspective also motivates new type of promising XAI approaches that are currently underexplored due to the largely intermingled view. In addition, I will showcase our AI Explainability 360 toolkit, which has diverse ways of explaining models and data along with educational material to guide folks that are new to this area.
00:00 Intro
03:28 WHY EXPLAINABLE AI ? Simplification
04:21 WHY EXPLAINABLE AI ? Debugging
05:22 WHY EXPLAINABLE AI ? Existence of Confounders
06:17 WHY EXPLAINABLE AI ? Enhance Performance
08:02 THREE DIMENSIONS OF EXPLAINABILITY
09:37 Post hoc (local) interpretation
13:42 ONE EXPLANATION DOES NOT FIT ALL
14:37 EXPLAINABLE AI: A 2-STAGE PROBLEM
16:37 EXPLAINABLE AI: POPULAR XAI METHODS
17:47 EXPLAINABLE AI: PROXY MODELS
20:57 EXPLAINABLE AI: MULTIHOP APPROACH
28:12 EXPLAINABLE AI: ONLY STAGE-1
32:37 AIX360: IBM RESEARCH AI EXPLAINABILITY 360 TOOLKIT (DEMO)
[Chapters were auto-generated using our proprietary software - contact us if you are interested in access to the software]
Lecture slides: https://drive.google.com/file/d/1S4Xe...
The presentation is based on the speaker's paper and project:
One Explanation Does Not Fit All: A Toolkit and Taxonomy of AI Explainability Techniques
https://arxiv.org/abs/1909.03012
https://github.com/Trusted-AI/AIX360
Presenter Bio:
Amit is a Principal Research Staff Member at IBM TJ Watson Research in NY. He has worked on projects spanning multiple industries such as Semi-conductor manufacturing, Oil and Gas, Procurement, Retail, Utilities, Airline, Health Care. His current research includes proposing various methods for enhancing trust in systems by developing methods that try to explain or understand their behaviors. His recent work was featured in Forbes, PC magazine, and NeurIPS. Besides research impact, his work has also gone into IBM product and he has received Outstanding Technical Achievement as well as IBM Corporate award. He has been an Area Chair and PC member for top AI conferences as well as has served on National Science Foundation (NSF) panels for the small business innovative research (SBIR) program. He also serves on the invention disclosure committee (IDT) in IBM Research.
-------------------------
Find us at:
Newsletter for updates about more events ➜ http://eepurl.com/gJ1t-D
Sub-reddit for discussions ➜ / 2d3dai
Discord server for, well, discord ➜ / discord
Blog ➜ https://2d3d.ai