MCCHE Precision Convergence Webinar Series with Gerd Gigerenzer
Homo Heuristicus: Decision Making under Uncertainty
By Gerd Gigerenzer
University of Potsdam
Date: February 6, 2025
Time: 11:00 AM to 1:00 PM
听
Abstract
In well-defined situations with known risks, the axioms of classical decision theory can guide optimal decision-making. However, when Savage introduced his axioms, he clarified that they apply to risk but not to uncertainty and intractability. Uncertainty refers to ill-defined situations where the future states of the world (their exhaustive and mutually exclusive set), their outcomes, and associated probabilities are either unknown or unknowable. Intractability, on the other hand, involves well-defined but overly complex situations, such as in games like chess or Go, where finding optimal solutions is impractical. Though Knight, Keynes, and Simon had drawn similar distinctions, most models of uncertainty have reduced it to risk, such as by using second-order probabilities, equal priors, or Bayesian subjective probabilities. In contrast, I argue for a genuine theory of decision-making under uncertainty, grounded in the empirical study of heuristic-based decisions. This approach includes three key research areas. The first is descriptive: What heuristics do individuals and organizations have in their adaptive toolbox, and how do they choose between them? The second is prescriptive: In what contexts are heuristics more likely to succeed than more complex strategies? This line of inquiry, known as the study of ecological rationality, examines the match between strategies (heuristics or others) and the structure of environments. The third area is engineering and intuitive design: How can we create heuristic systems that aid experts and non-experts in making better decisions? To achieve this, three methodological tools are essential: formal models of heuristics (to go beyond vague terms like "System 1"), competitive testing of heuristics against complex strategies (instead of merely relying on null hypothesis testing), and evaluating the predictive power of heuristics (rather than just fitting them to data). Through examples from finance, management, and sports, I demonstrate that heuristics often predict as accurately, if not better, than complex strategies, including some machine learning algorithms.
