The Chemical Engineering Department is hosting a special graduate lecture on AI Transparency & AI Transformative Utilizations in Chemical Engineering.
Biographical Sketch
Donovan Chaffart is an assistant professor at the Eastern Institute of Technology (EIT)’s College of Engineering in Ningbo, China. Originating from Canada, he obtained his PhD in Chemical Engineering from the University of Waterloo, before recently joining EIT as a professor. In addition, he has established himself in the development of multiscale and reliable AI systems for Chemical Engineering applications—including interfacial systems, fluidization, and film deposition. Despite only just beginning his academic career at the EIT, Dr. Chaffart has already co-founded the Intelligent Multiscale Process Engineering research group along with Dr. Yue Yuan. Within this research group, he continues his pursuit in developing and implementing novel numerical modelling solutions, high-quality multiscale simulations, and transparent AI—all for the objective of tackling and solving critical problems within the field of Engineering. As a result, Dr. Chaffart has rapidly proven himself over his academic career to be an expert in both multiscale and AI-assisted modelling. Throughout his young academic career, the applicant published over 15 articles within high impact factor journals, including 10 papers as a first or corresponding author. Thanks to his specialisation in AI, Dr. Chaffart has been invited to give presentations on his research on transparent AI at both the International Metallurgical Materials and Chemical Engineering Intersection Symposium held in July of 2023 in Baotou, Inner Mongolia, and at the Artificial Intelligence Research and Applications in Chemical Engineering conference held in Ningbo, Zhejiang in August of 2023. Most recently, he was invited to be a keynote speaker for the 2024 Artificial Intelligence Research and Applications in Chemical Engineering conference.
Abstract:
The last few years have seen one of the biggest growths in Artificial Intelligence (AI) and have seen its rapid growth and adoption within a wide range of different industries. However, AI has seen relatively slow growth within sensitive fields concerning health, safety, and high profitability, such as Chemical Engineering. This is primarily due to the black box nature of modern data-driven AI methods, which limit both their capacity to justify their outputs and our own capacity to understand the underlying rationales of a trained AI model. This talk will discuss the concept of Explainable AI (XAI) and how it can be used to improve the reliability of AI models and thus improve our confidence in developed models within the Chemical Engineering industry. This talk will discuss the requirements of reliable AI regarding the transparency of its transformative utilization within CE from the aspects of causality, explainability, and informativeness, and the current techniques that have been developed to address these concerns. Each of these XAI aspects will be provided metrics that can be used to assess the underlying transparency embedded within a trained data-driven AI model. Furthermore, this talk will discuss current setbacks within each of these aspects for Chemical Engineering applications and will provide preliminary solutions that are being investigated by Dr. Donovan Chaffart’s research group in order to develop transparent, reliable AI methodologies for use in critical domains such as Chemical Engineering.