Benjamin Recht, The Irrational Decision. Princeton University Press. 2026. ISBN13: 978-0-691-27244-3.
In The Irrational Decision: How We Gave Computers the Power to Choose for Us from Princeton University Press, Benjamin Recht of UC Berkeley examines the history of computer-driven risk analysis from its foundations in World War II to the present day. Many of Dr. Recht’s themes are familiar, but he provides ample new (to me, at least) perspectives and examples to advance the discussion.
I particularly like his treatment of creating linear models, which have well-defined solutions, as a means of controlling physical systems. Another recent book from Princeton, Disneyland and the Rise of Automation (which I reviewed here) examines automation as applied to theme park ride control and optimization. Recht acknowledges the power of these systems but references a 1989 conference plenary address entitled “Respect the Unstable” by Gunter Stein, a Honeywell engineer, regarding the dangers of forcing linear control over nonlinear systems. Recht notes that Stein provided several examples of small perturbations that caused a nonlinear system controlled using a linear model to get dangerously out of control.
Other examples include gameplay for checkers, chess, go, and poker. Checkers, chess, and go are perfect information games with enormous game trees, while poker has a huge game space that includes hidden information. As a risk analysis and optimization problem, it’s theoretically possible to describe the entire game tree of a perfect information game. That has been done for checkers, which was solved in 2007, but seems comfortably out of reach for chess and go.
It’s different for business problems. You can’t build a decision tree or evaluate simulation results without knowing the specific outcome of each play of the game. Defining player payoffs in all but the simplest cases is impossible, especially if you include non-monetary measures. To that end, Recht quotes RAND historian Robert Leonard:
Repeatedly, one encounters the opinion that, as far as systems analysis was concerned, the value of game theory at RAND was not in providing numerical solutions to particular strategic problems, but in providing qualitative insight, in stressing the important of reciprocation, opposing intentions, the credibility of threats, and the like. Game theory played a role in systems analysis, therefore, but not as a generator of numerical solutions to well-specified problems.
How, then, can and should humans make decisions? The Irrational Decision compares the competing theories of Daniel Kahneman of Thinking Fast and Slow fame, and Gary Klein, who wrote Sources of Power and Streetlights and Shadows. Kahneman focuses on heuristics and rules, while Klein created his Recognition-Primed Decision Making (RDM) model. Recht’s comparison of the two approaches and how the creators achieved a form of reconciliation was interesting to read and more information I hadn’t encountered previously.
The Irrational Game is an excellent book. Even if you’re familiar with the popular literature on decision making you’ll likely find new information and fresh perspectives. Highly recommended.