jessi_cata and OrionJohnston Debate Whether FDT's Dependency Hypotheses Are Truly "Logical"
On November 10, jessicata and OrionJohnston held a multi-round technical debate on X over functional decision theory (FDT), centered on FDT's terminology and theoretical core.
Confirmed
- OrionJohnston argued that FDT's dependency hypotheses are essentially "logical": the causal graph contains logical nodes, interventions act on these nodes, and this captures cross-world dependence on logical facts—a key way to distinguish FDT from CDT.
- jessicata countered that the original FDT paper itself conflates "logical" dependency hypotheses, so the paper should be criticized directly and alternative theories given new names, rather than relying on what she considers an untenable notion like the "FDT family."
- jessicata noted that under OrionJohnston's taxonomy, "physical CDT" is actually closer to the mainstream version and has a concrete constructible form—for example, connecting a quantum random number generator to a computer running an RL-like algorithm, using the RNG for random exploration, which also matches intuitions about "physical causation." In her view, this exposes FDT's lack of an equivalent concrete construction.
- The two also clashed over the meaning of "causal": jessicata argued for distinguishing physical causation from other uses such as non-backtracking.
- jessicata further pointed out that assumptions like "my choice determines both my own action and Omega's" are not logical-level facts but the environment's dependence on one's policy, and should not be conflated with FDT's logical character.
Why it matters
- The debate strikes at FDT's theoretical foundations: whether the "logical" nature of dependency hypotheses holds determines whether the boundary between FDT and the CDT family is well-defined.
- The quantum RNG + RL exploration paradigm gives CDT an operational, concrete construction, highlighting the methodological gap between "theories with constructive forms" and "theories resting on intuitive descriptions"—an ongoing concern in decision theory circles.
2026-11-10 ~ 2026-11-10 · 10 related posts
Primary sources
- [source] Physical CDT has a constructible form; FDT has none — an asymmetry — jessi_cata · 2026-11-10
- FDT Defenders Push Back on "Logical Dependency Hypotheses Aren't Really Logical" — jessi_cata · 2026-11-10
- FDT debate: are dependency hypotheses 'logical'? Two researchers clash — OrionJohnston · 2026-11-10
- FDT Debate: jessi_cata Argues FDT Is Logical, Policy Dependence Is a Different Concept — jessi_cata · 2026-11-10
- [source] OrionJohnston: FDT Dependency Hypotheses Are a Free Parameter Known Only via Commentary — OrionJohnston · 2026-11-10
- jessi_cata: FDT Paper's Verbal Commentary Is Part of the Theory, Not Just Equation 4 — jessi_cata · 2026-11-10
- [source] FDT Debate: Is the 'Logical' Label a Mistake? Decision Theory Iterates in Public — jessi_cata · 2026-11-10
- Naming a Successor to FDT: The 'Logical' Mislabel Fight Goes On — OrionJohnston · 2026-11-10
- CDT paradigm case proposed: quantum RNG hooked to RL-style randomized exploration — jessi_cata · 2026-11-10
1 near-duplicate retellings: jessi_cata