On Monday (10/7), my coauthor Justin Klevs and I presented processes and examples for explainable AI at the Bank AI Chicago conference. It was also an opportunity to listen to and interact with bankers, regulators, data scientists, and technologists touching upon various topics in operationalizing AI, bias in AI, privacy, interpretability and explainability. One important takeaway is that explainability should be designed to address multiple stakeholder concerns, especially in a highly regulated industry.
On Friday (10/11), there was an interactive discussion with data scientists and technologists from multiple industries at the Reston Azure DataFest. It was also an opportunity to demonstrate our explainable AI platform. Thanks to my colleague Vinh Nguyen for showing multiple examples on instance level prediction and overall prediction explanations, including feature interactions. Below example shows an actionable benefit.