AI and Organizational Decision Making

Complexity Science provides a crucial lens for considering how AI will impact human systems. In this session, SFI External Professor Scott Page* presented some ideas from a current project with Lu Hong and Anusha Kallapur exploring the potential role (and limitations) of AI in generating novel ideas and decisions on large strategic decisions in complex settings.

They posited that if AI adds meaningful value to strategic decisions, then it must produce "surprises,” or insights that differ from what human decision-makers would entertain. However, when these surprises occur, organizations may not be able determine whether the AI is (1) correct and using novel reasoning that adds value, (2) hallucinating by identifying false patterns in data, or (3) misaligned with the organization's actual goals. This “Inferential Trilemma” is especially problematic for high-stakes strategic decisions like mergers, R&D investments, or market entry, where outcomes are uncertain and AI reasoning may remain opaque due to its reliance on massive datasets with billions of parameters.  Even if ex ante, the probability of misalignment and hallucination are low, conditional on a surprise that potential adds high value, the probability may be large.

*Scott also serves as the John Seely Brown Distinguished University Professor of Complexity, Social Science, and Management at the University of Michigan Ross School of Business and Professor of Complex Systems, Political Science, and Economics in the College of Arts and Sciences and as a non resident fellow at the Brookings Institute.