Philosophy

Argues that Bayesian frameworks dissolve rather than solve traditional philosophical problems, including Gettier cases and Hume's problem of induction. A systematic application of probabilistic reasoning to longstanding epistemological debates.

Argues that the Bayesian polarization literature asked the wrong question. "How can rational agents diverge given common evidence?" presupposes common evidence — but partisan media ecosystems violate this premise at scale. The convergence theorems fail not because of subtle cognitive mechanisms but because their central assumption is empirically false. Twenty years of cognitive interventions failed because they target the wrong failure mode.

Proposes updateability — the capacity of a theory to be meaningfully revised in light of new evidence — as a criterion for demarcating science from non-science, offering an alternative to falsifiability.

Goodman's "new riddle of induction" asks why we project "green" rather than the gerrymandered "grue." This paper argues the asymmetry is physical, not linguistic: projectible predicates are those preserved under the symmetries of physical law, so the green/grue distinction follows from dynamics rather than convention. Developed as a standalone argument split off from the demarcation paper, where a referee called it "very much original and worthy of consideration."

Dissolves the Boltzmann brain hypothesis through information propagability and Liouville's theorem. The argument has two horns: ex-nihilo brain formation is forbidden by symplectic structure; Poincaré-recurrence variants are philosophically moot because they cannot ground the kind of skeptical hypothesis the literature requires.

The explicit Turing test was built for the wrong question — the imitation game presupposes the very consciousness it was meant to detect. The paper proposes the implicit Turing test instead: a system is intelligent iff it can detect that its own constitutive interpretive frameworks have failed and revise them under world-pressure. Gödel's incompleteness, Turing's halting problem, and the Löbian obstacle are shown to be one diagonal argument — no sufficiently rich system can justify its own framework from within, so external grounding is the only resolution that does not relocate the regress. Current AI is mapped onto a five-level taxonomy of change, with the most advanced self-improving architecture bounded at Level 3a by the Löbian obstacle, not by engineering.

Applies graduated Bayesian credences about moral status to generate graduated protections, offering a framework that dissolves the binary framing of the abortion debate.

Develops ethical principles for consciousness research on AI systems using Talmudic legal reasoning. Addresses the moral paradox of investigating consciousness in entities whose moral status depends on the investigation's outcome.