Title: "Understanding Human Trust in Machine Learning through Behavioral Experiments"

Abstract

Machine learning (ML) has shown tremendous potential in transforming numerous industries and nearly every aspect of our daily life. Today, innovative ML techniques have been developed and applied to various domains from finance, to healthcare, to criminal justice, and they achieve great success in uncovering hidden insights from massive data and improving the outcomes of decision making. Despite the promise, the human-facing nature of modern ML systems highlights the needs of placing humans in the central role in designing and evaluating ML systems and raises fundamental new challenges on understanding how people interact with and trust these systems. In this talk, I'll discuss a series of studies, in which we design and conduct human-subject randomized behavioral experiments to understand how people's trust in ML models is influenced by a variety of factors, including the model's stated accuracy on held-out data, the model's accuracy observed in practice, and the model's confidence on individual predictions. I'll conclude by discussing the implications of our experimental results on designing better ML systems.

Bio

Ming Yin is an assistant professor in the Department of Computer Science, Purdue University. Her research broadly connects to the fields of human-computer interaction, applied artificial intelligence and machine learning, computational social science, and behavioral sciences. Ming's research focuses on using both experimental and computational approaches to examine how to better utilize the wisdom of crowd to enhance machine intelligence (e.g., crowdsourcing), and how to better design intelligent systems that people can understand, trust and engage with effectively (e.g., human-AI interaction). Ming is named as a Siebel Scholar (Class of 2017), and she has received Best Paper Honorable Mention Award at the ACM Conference on Human Factors in Computing Systems (CHI), in 2016 and 2019.

Prior to Purdue, Ming spent a year at Microsoft Research New York City as a postdoctoral researcher in the Computational Social Science group. Ming completed her Ph.D. in Computer Science at Harvard University, and received her bachelor degree from Tsinghua University, Beijing, China.