Abstract:
If ‘Data is the new Oil’, AI is the refinery where usable products are distilled.
There is a growing trend towards digitalization and an ever-increasing reliance on machine learning based tools for decision making. Deep learning is seen as the pre-eminent approach not only to classify photos of dogs and cats, but also to decide the credit card limits, pain medication dosage, and even the policing level to be deployed in a locality. In manufacturing, this trend, often christened as Industry 4.0, is most widely seen in predictive maintenance, inventory management, robotics, and quality management. As the adoption of AI becomes prevalent at scale, it is becoming clear that deep learning models do not inherently possess good nterpretability, i.e., the end-user is unable to understand why a model has reached a certain conclusion. With the democratization of AI, sophisticated models can be developed using just drag-and-drop by a non-AI-expert. Transparency of these models is therefore more important than ever. These issues are now being addressed by eXplainable Artificial Intelligence (XAI) – a new field in machine learning that seeks to provide tools and techniques that enable human users to comprehend and trust an AI model.
In this seminar, I will provide a panoramic view of XAI from the view of process systems engineering. We will discuss the forms of explanations, their conceptual underpinning, and the various stages they can be useful. We will also review how the benefits of a transparent AI model can be quantified and the cognitive value of an explanation. Our research group studies the interaction of human cognition and complex decision making, especially in high-risk domains. Using examples from our studies, I will illustrate the methods of XAI as relevant to PSE. Some current challenges and future research directions will also be discussed.
Speaker Bio:
Rajagopalan Srinivasan is a Professor of Chemical Engineering and the Head of the American Express Lab for Data Analytics, Risk & Technology (DART Lab) at Indian Institute of Technology Madras. Raj received his B.Tech from IIT Madras and PhD from Purdue University.
Raj’s research program is targeted towards developing systems engineering approaches for the design and resilient operation of complex systems. He is an author of over 400 peer- reviewed journal and conference articles, book chapters and technical reports. He has received Best Paper Awards from several journals & conferences and is ranked among the Top 2% scientists in the category of Chemical Engineering. He is a consultant to over 20 well-known companies such as ABB, ExxonMobil, Honeywell, and Shell. He is an Associate Editor for IChemE’s Process Safety and Environmental Protection, Frontiers in Energy Research, and PLOS One and sits on the Editorial board of over 20 international journals in his field.