Episode Summary
Executive Summary: Christian Tarsney of Oxford’s Global Priorities Institute explains why future bias, fanaticism, and epistemic uncertainty matter for deciding how to do the most good. He argues that long-termist and expected-value views are more nuanced than critics claim, and that background uncertainty can justify action without requiring extreme Pascalian fanaticism.
Main Topics: Global Priorities Research and the far future: Tarsney describes GPI’s mission: building philosophy/economics tools to help altruists prioritize among causes, especially those affecting the far future, where empirical prediction is hard and uncertainty is high. Future bias and the value of past vs future experiences: The discussion explains why people typically care more about future pain than past pain, why that may stem partly from inability to affect the past, and why that doesn’t fully justify temporal asymmetry normatively. Philosophy of time and time travel intuitions: Rob and Christian review presentism, growing block theory, eternalism, and moving spotlight views, and how time-travel or retrocausal scenarios complicate intuitions about whether the past matters. Decision theory, Newcomb’s problem, and past-affecting choices: Christian connects future bias to evidential vs causal decision theory, arguing that in some cases choices that alter past experiences or signal information can affect rational action even without literal time travel. Fanaticism and Pascal’s mugging in expected value reasoning: The episode centers on whether tiny probabilities of enormous outcomes should dominate decisions. Christian argues that standard expected utility does not force linear maximization and that stochastic dominance plus background uncertainty offers a middle path. Epistemic challenge to long-termism: Christian’s long-termism work models how predictability decays over time and whether existential risk reduction still looks worthwhile under pessimistic assumptions about persistence and future stability. Moral uncertainty and externalism vs internalism: The conversation ends by discussing whether uncertainty about moral theories should affect decisions, how to avoid regress over higher-order moral uncertainty, and a moderate externalist position about rationality.
Key Arguments: People often care more about future than past experiences because the past seems unaffectable, but that explains only part of the intuition and does not by itself justify temporal bias. Experimental work on hypothetical time-travel cases suggests that when people can affect their past experiences, future bias weakens but does not disappear. Expected utility theory does not force linear scaling of value; it only requires a utility ranking plus probabilistic aggregation, so very concave utility functions can avoid some fanaticism. A minimal stochastic-dominance standard, combined with background uncertainty about the universe’s baseline value, can recover much of ordinary expected-value reasoning without endorsing all Pascalian extremes. Long-termist choices can remain highly valuable even under pessimistic assumptions about decay of influence, because even small probabilities of persistent future success can have enormous expected value. Moral uncertainty should matter, but uncertainty about the very principles of rationality should not create an infinite regress; a moderate externalism about rationality avoids that problem. In practice, much of what looks like long-termism may actually resemble common-sense priorities such as pandemic prevention, nuclear risk reduction, institutional quality, and better governance.
Data Points: Podcast scope: ~2+ hours of philosophical discussion - The transcript covers several linked topics: future bias, decision theory, fanaticism, long-termism, and moral uncertainty. Probability example: 1 in a Google (10^-100) - Used to illustrate an extremely tiny chance of an astronomically good outcome in the fanaticism discussion. Survival impact assumption: At least 1% - Christian’s back-of-the-envelope estimate of how much civilization-focused effort might improve humanity’s survival probability over a millennium. Annual existential catastrophe rate (pessimistic assumption): 1% per year - Presented as an extreme pessimistic assumption in the long-termism model. Alternative annual existential catastrophe rates: 1 in 10,000 to 1 in 1,000,000 - Examples of more optimistic rates that, if given even small credence, can restore strong expected-value support for long-termism. Potential scale of future value: 10^52 happy lives - Illustrative estimate for the value of a vast future civilization in the fanaticism discussion. Fanaticism threshold claim: Roughly out to probabilities like 1 in a billion - Christian suggests background uncertainty may make expected-value-like reasoning compelling for probabilities above very tiny levels. Debate speech rate: 300–350 words per minute - Describes the very fast pace common in U.S. competitive debate rounds. Time horizon example: 1,000 years - Used repeatedly as a horizon for far-future prediction and existential risk modeling.
Pivotal Quotes: "There are good reasons to be skeptical of that." — Christian Tarzany: On the idea that naive expected-value maximization is automatically required. "What we should really be doing is, for instance, trying to make existing societies more fair and just and equal in order to improve social trust and social capital." — Christian Tarzany: Describing a subtle form of long-termism focused on present institutions that shape the future. "The question we're asking here is: what is the rational thing to do under uncertainty in general?" — Christian Tarzany: Summing up his view on moral uncertainty and the scope of rational decision theory.
Implications: Listeners should come away less certain that “maximize expected value at all costs” is the only rational approach. The episode suggests a practical middle ground: prioritize robustly important long-term risks, but use uncertainty-aware standards rather than extreme Pascalian fanaticism.