96. Brian Skyrms | Decision Theory

When should evidence guide choice over causation, how can meaning emerge from signaling games, and why might utility comparisons quietly break the ethics we build on them?My links: https://linktr.ee/frictionphilosophy.1. GuestBrian Skyrms is Distinguished Professor of Logic and Philosophy of Science and Economics at the University of California, Irvine, and an Emiritus Professor of Philosophy at Stanford University. His work has focused on science, causation, decision theory, game theory, and the foundations of probability.2. Interview SummaryBrian Skyrms frames the opening around the long-running split between causal and evidential decision theory: both aim at “maximizing expected payoff,” but they compute “expectation” using different probability functions—ordinary degrees of belief for the evidentialist, versus probabilities meant to represent causal efficacy for the causal theorist. He suggests that once you’re clear on that conceptual difference, much of the remaining literature becomes a matter of pushing (and disputing) intuitions about “funny cases,” especially Richard Jeffrey–style “news value” reasoning versus causal evaluation in cases like Newcomb-style problems, which he’s surprised to see repeatedly “rise from the grave” in new philosophical (and AI-adjacent) waves of debate.From there, the interview broadens into Skyrms’s broader picture of how decision theory connects to a cluster of topics—conditionals, causation, and modal/necessity talk—where he favors a pragmatic, human-centered treatment rather than hunting for extra “facts” supposedly delivered by raw intuition. In that spirit, he discusses Bayesian-friendly ways of handling subjunctive conditionals that preserve some “probability of a conditional” motivations while avoiding classic triviality worries, and he emphasizes that what counts as “rational” can shift with the setting: in correlated evolutionary contexts (where you reliably meet similar agents), evidential/Jeffrey-style reasoning can predict what evolves better than individualistic causal-choice reasoning, which helps motivate his interest in the evolution of cooperation.In the second half, Skyrms highlights two big research programs. First, he explains how a naturalistic account of meaning can start from information transfer in David Lewis–style signaling games—tracking how learning dynamics can move populations toward equilibria where signals stabilize—and then treating deception as a deviation (strategic or mistaken) from an emergent equilibrium use-pattern. Second, he summarizes his recent work on utilitarianism as largely a measurement-theory project: many familiar philosophical “add up the utils” arguments implicitly assume overly strong scales, and once you respect the legitimate degrees of freedom in utility measurement, celebrated cases (including Derek Parfit-style population examples) can become formally meaningless because permissible rescalings flip the verdicts. He then connects this methodological moral to epistemology: you can’t “answer” the complete skeptic (that’s a fool’s game), but you can make progress with partial skeptics by getting precise about what “the future will be like the past” could even mean—a point associated with Nelson Goodman—and by using Bayesian tools to articulate which inductive inferences are actually supported. He closes on a deflationary note about philosophy’s value: he won’t sell it as for everyone; for him, it’s mainly that it’s fun to question what people accept without thinking.3. Interview Chapters00:00 - Introduction01:19 - Causal and evidential decision theory04:05 - Maximizing expected payoff05:27 - Correlation without causal connection06:17 - Causation10:41 - Deliberating over the past13:42 - Common intuitions16:11 - Philosophical landscape18:05 - Predicted riches21:27 - Transparent box case23:55 - Normative significance of decision theory28:33 - Subjunctive conditionals33:33 - Closest possible worlds35:15 - Backtracking counterfactuals37:07 - Example39:47 - Possible worlds41:38 - Metaphysical possibility44:05 - Suppositional approach48:10 - Benefits of the approach51:19 - Foundations of utility56:55 - Utilitarianism and measuring utility1:01:28 - Saving utilitarianism1:02:43 - Vague preferences and credences1:06:10 - Content in signalling games1:13:48 - Not sui generis1:15:23 - Signalling games with only one person1:18:18 - Accounting for content more broadly1:21:01 - Inductive skepticism1:23:50 - Required assumptions1:25:55 - Problem of induction1:31:11 - Contracts and games1:33:11 - Correlations1:38:33 - Value of philosophy1:39:56 - Conclusion This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit fric.substack.com/subscribe