Pragmatic FDT, and predictors as game theory

Functional decision theory is having a renaissance—or so its advocates say. According to LessWrong, decision theorist Stuart Armstrong argues that functional decision theory, despite its theoretical problems, can be salvaged. Rather than getting tangled in abstract questions about whether two algorithms are truly identical, he proposes building concrete evidence of equivalence instead. His method, called pragmatic FDT, works in four steps: start with a baseline rational decision, search for likely isomorphisms—structural similarities between your decision process and the world around you—evaluate each one, and adopt the best that actually improves your outcome. The key insight is elegant: two calculators don't need to be philosophically identical to behave the same way. A calculator that outputs negative numbers is isomorphic to a standard one if you relabel the negatives. Armstrong also connects this to game theory: whenever a predictor makes counterfactual guesses about your choices, decision theory becomes strategic negotiation rather than pure reasoning.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton