Human and machine learning: the search for anomalies

Machine learning is perhaps the most significant field in computer science now, and is showing dramatic improvements in capabilities, wide-spread uses, but also confusing limitations. This project seeks to address limitations in machine learning and explores the question: what is the nature of tasks that are easy for humans to learn but hard for machines, and vice versa?

Understanding differences in how humans and machines learn has the potential both to greatly improve machine learning, if it can be adapted to be more human, and also to build systems that optimally blend humans and machines in ways that best leverage their inherent strengths.

Gaining such knowledge will generate new insights into human problem-solving, and will help us understand what kind of prior learning is necessary for improving machine learning. Finally, the project will explore the potential for translation to real-world operational challenges. If we know when humans learn better, and when machines do, can we triage learning problems, or establish better partnerships between the two?

PRINCIPAL INVESTIGATOR
Vicki Bier, professor of industrial and systems engineering

CO-PRINCIPAL INVESTIGATORS
Gary Lupyan, associate professor of psychology

Xiaojin Zhu, professor of computer sciences

Paul Kantor, honorary associate, industrial and systems engineering