Gerard Joseph Rothfus

Consider two principles of rational decision making. One says that you should never use the value of events that are inconsistent with your evidence to assess what to do. The other says that, if you know what a more informed version of yourself would do, then you should already do likewise with your current information. We show that no decision theory which agrees with evidential decision theory (EDT) and causal decision theory (CDT) whenever they agree can satisfy both principles. In particular, assuming that agents update via Bayesian conditioning, EDT satisfies the former and violates the latter, while CDT violates the former and satisfies the latter. We show that the causalist can satisfy the first (at the cost of then violating the second) if she either uses a learning rule that changes her beliefs about counterfactuals in a particular way or uses a decision rule more in line with a rule proposed by Hitchcock (2016). We then employ these results to illuminate a core tension at the heart of the debate between evidentialists and causalists.

Trolley problems and like cases are often thought to show the inadequacy of purely consequentialist moral theories. In particular, they are often taken to reveal that consequentialists unduly neglect the moral significance of the causal structure of decision problems. To precisify such critiques and one sort of deontological morality they motivate, I develop a formal modeling framework within which trolley problems can be represented as suitably supplemented structural causal models and various consequentialist and double effect-inspired moral theories can be viewed as disagreeing over the inputs of a common decision rule.

Philosophers’ two favorite accounts of rational choice, Evidential Decision Theory (EDT) and Causal Decision Theory (CDT), each face a number of serious objections. Especially troubling are the recent charges that these theories are dynamically inconsistent. I note here that, under the epistemic assumptions that validate these charges, every decision theory that satisfies a pair of attractive postulates is doomed to a similar fate and then survey various lessons rational choice theorists might opt to draw from this.

In ‘An argument against causal decision theory’, Jack Spencer shows that standard formulations of causal decision theory run afoul of his Guaranteed Principle. In the sequential choice problem he employs to make this case, the transgression stems from an awkward discrepancy between how causalists typically value present vs future acts. This note suggests a version of causal decision theory that avoids this incongruity and so respects the Guaranteed Principle in Spencer’s problem. However, this formulation, and hence symmetric appraisal of present and future acts, is also shown to be insufficient to secure causalist satisfaction of the Guaranteed Principle in general.

Arif Ahmed has recently argued that causal decision theory is dynamically inconsistent and that we should therefore prefer evidential decision theory. However, the principal formulation of the evidential theory, Richard Jeffrey’s Logic of Decision, has a mixed record of its own when it comes to evaluating plans consistently across time. This note probes that neglected record, establishing the dynamic consistency of evidential decision theory within a restricted class of problems but then illustrating how evidentialists can fall into sequential incoherence outside of this class. Uncovering the evidentialist’s own dynamic inconsistencies reveals, contra Ahmed, that sequential choice considerations do not significantly favor the evidentialist’s theory over the causalist’s.

One of the main contributions of Richard Bradley’s book is an elegant extension of Jeffrey’s Logic of Decision that countenances the evaluation of conditional prospects. This extension offers a promising new setting in which to model dynamic choice. In Bradley’s framework, plans can be understood as conditionals of an appropriate sort, while dynamic consistency can be viewed as providing a constraint on the evaluation of conditionals across time. In this paper, we study connections between planning conditionals and dynamic consistency.