84. Paul Weirich | Decision Theory, Rationality

Can probability really be a “guide to life” for individuals and groups once we take seriously risk, coordination limits, and imprecise beliefs?

My links: https://linktr.ee/frictionphilosophy.

1. Guest

Paul Weirich is Curators’ Distinguished Professor Emeritus in philosophy at the University of Missouri, and his work has focused primarily on decision theory, game theory, rationality, and logic.

2. Interview Summary

Paul Weirich Paul Weirich frames the discussion around a broadly Bayesian ideal—“probability as a guide to life”—and then uses that outlook to motivate a distinctive approach to collective rationality. Bishop Butler is invoked as the source of the “guide to life” slogan, and the thought is that advice gets sharper when it’s made explicitly probabilistic. From there, Weirich argues that groups really can act (a committee can pass a resolution) even if they don’t literally have beliefs, desires, or intentions; because they act freely and without coercion, their actions can still be assessed by standards of rationality. But the standards for groups differ from those for individuals: since collective action is realized only through the members’ actions, he proposes a “compositional” sufficient condition—if the individual actions that constitute the group act are rational, that suffices for the group act’s rationality. He also pushes back on the idea that collective rationality must meet demanding criteria like Pareto optimality, emphasizing that real-world limits on coordination and communication can lower what it’s fair to demand.

The interview then turns to how decision theory should represent what agents care about—especially risk, cooperation, and time. Weirich defends the expected-utility framework but argues that the “possible outcomes” used in calculating expected utility should be comprehensive: they should cover everything the agent cares about, including risk (for agents who care about it). He distinguishes “local” risk (a chance of a bad outcome), toward which rationality requires aversion, from “global” risk (riskiness as a trait of a life or plan), toward which rationality may permit much more tolerance—even attraction. Relatedly, in discussing familiar puzzles about correlation and choice (including Newcomb-style setups), he contrasts evidential and causal decision theory and explains why, as a causal decision theorist, he takes the point of choosing to be promoting one’s goals rather than merely producing a good sign of goal-promotion. That segues into sequential choice and planning: because you can’t “presto” perform all steps of a multi-stage course of action at once, he thinks sequences should also be evaluated compositionally—by assessing whether each step is rational at the time it’s performed—rather than by asking only whether the whole sequence maximizes utility among rival sequences.

Finally, Weirich develops his view about imprecise credences and utilities. He argues that in some circumstances—especially when information is scarce—rationality may not merely permit but even require imprecise probabilities, and this forces a generalization of the usual “maximize expected utility” rule. On his preferred generalization (associated with work going back to I. J. Good), an agent’s state of mind is represented by a set of probability/utility pairs, and rational choice can be understood permissively: choose an act that maximizes expected utility relative to some admissible pair in that set. He resists proposals that collapse imprecision into a single “median” credence, insisting that genuine imprecision should sometimes change what it’s rational to do. He then connects this to a defense of the probability axioms: rather than relying only on Dutch-book style pragmatic pressure, he favors a coherence-based rationale that (he claims) extends cleanly to the imprecise case, and he treats these coherence constraints as non-instrumental norms of rationality rather than mere tools for getting what one happens to value.

3. Interview Chapters

00:00 - Introduction

01:14 - Group rationality

04:27 - Compositionality of rationality

07:12 - Rational groups with irrational members

12:53 - Each/we dilemmas

18:09 - Prisoner’s dilemma

20:17 - Risk and decision theory

23:48 - What is risk?

27:45 - Risk as consequences

31:18 - Risk as built into preferences?

34:50 - Non-linear preferences

37:30 - No constraints?

40:03 - Imprecise credence

43:57 - Modelling credence

46:28 - Permissive rationality

47:34 - Preferring median representatives?

53:11 - Vagueness

56:17 - Intrinsic epistemic goodness

1:00:35 - Instrumental vs. non-instrumental

1:02:34 - Value of rationality

1:05:42 - Rewarded irrationality

1:09:24 - Newcomb’s problem

1:19:45 - Sequential decision problems

1:22:18 - Self-torturer paradox

1:25:40 - Rationality without binding?

1:30:08 - Value of philosophy

1:32:08 - Conclusion