Biases of Humans, of AI, and of Humans with AI
Abstract
A large body of behavioral science explores decision biases of humans, many of which are of direct relevance to managerial decision making, including in operational contexts. With the recent advances in artificial intelligence (AI), however, for many managerial decisions people obtain inputs from AI or even fully delegate decisions to AI. Within this reality, this talk will discuss two related papers. The first one sources the list of biases particularly relevant for business operations, and explores them with the currently dominant AI, ChatGPT. One of the main findings from this paper is that three patterns are observed: AI may be biased in the same way as humans, it may be rational (that is, exhibit no bias, which is different from humans), or it may be biased differently (that is, be neither human-like nor rational). More so, these patterns are generally consistent across contexts, with some variation across ChatGPT models and versions. The second paper focuses on one bias from each pattern and explores how humans change their behaviors (or not) when interacting with the AI that may be biased as a human, unbiased, or biased differently. The paper also makes a methodological contribution by developing a novel experimental design paradigm that allows to capture rich subject-driven AI interactions beyond traditional interfaces where the nature of interactions is pre-determined by the experimenters.
About the speaker
Anton Ovchinnikov is an Associate Professor of Decision Sciences at INSEAD. He holds a specialist degree in Economics from his hometown university in Krasnoyarsk, Russia, and a PhD in Operations Management from the University of Toronto. Prior to INSEAD he was a professor at the Darden School of Business at the University of Virginia in the USA and at the Smith School of Business at Queen’s University in Canada. Under his leadership, the Smith School won a major global prize for the best curriculum in analytics and AI: the UPS Prize by the Institute of Operations Research and Management Sciences (INFORMS). Anton’s research interests are on the intersection of Decisions Sciences with Operations, Marketing, and Economics. On the theoretical side these include studying strategic behavior of consumers and firms, impacts of innovative technologies, and environmental and social sustainability, and on the applied – data driven decision making in business, government, and non-profits. Anton published extensively in both the leading academic research journals, such as Management Science (MS), Operations Research (OR), Manufacturing & Services Operations Management (MSOM) and Production and Operations Management (POM), and the leading practice-oriented journals, such as Harvard Business Review (HBR).
Language
English
This is a free seminar. Registration is mandatory.