Axio › Volume 2 — Conditionalism: Truth, Bayes, and Rationality

Volume 2 — Conditionalism: Truth, Bayes, and Rationality

Isaiah Berlin sorted thinkers into hedgehogs, who see everything through one big idea, and foxes, who distrust big ideas and know many small things. The hedgehog risks dogmatism — complexity forced into a system too neat to be true. The fox risks paralysis — nuance entertained forever, with no framework sturdy enough to act on. This volume is written in a third posture: one coherent framework, held the way a fox would hold it — conditionally, humbly, with every assumption on the table and open to attack.

The framework is Conditionalism: the thesis that all truth is conditional — that every claim bears its truth value only relative to background conditions, stated or hidden, and that progress in thinking is largely the work of dragging those conditions into the light. That single idea, followed honestly, turns out to reorganize everything epistemology cares about: what truth is, what statements mean, what models and beliefs are, what knowledge amounts to, how probability works, and what it is rational to do when you cannot be certain — which is always.

The argument runs in six movements. Part I derives philosophy itself from the situation of any bounded agent that must act on incomplete information, and locates philosophy’s job: making pre-theoretic commitments explicit and testing them for coherence. Part II develops the theory of conditional truth — against absolutism and relativism alike — through the three levels at which truth operates, the ways statements fail to say anything at all, and the twentieth-century formal results that arrived at the same verdict from inside logic itself. Part III turns to representation: all empirical knowledge is model-mediated, and realism survives that fact in conditional form. Part IV rebuilds the theory of mind’s epistemic furniture — belief as a modeling construct, knowledge as reliable entropy reduction, and faith as the one epistemic structure the framework must condemn: belief with its update rule frozen shut.

Part V is the volume’s technical heart. It separates the two things ordinary language calls probability — Measure, the objective weight of branches in a quantum universe that never collapses, and Credence, an agent’s rational uncertainty about where in that structure it stands — and shows how Bayesian updating couples them, why the critical rationalists’ attack on Bayes fails, what varieties of uncertainty an agent can coherently carry, and why you are neither a random sample among observers nor a random branch among worlds. Part VI brings the machinery back to practice: rationality without foundations, coherence as the one value that can be held sacred without freezing thought, the discipline of changing your mind, the diagnosis of persistent disagreement, a live calibration exercise on a question your tribe has already answered for you, the discipline of knowing when a probability has been earned at all — the frame precedes the number — and decision-making when the tails are fat and the futures are many. A coda gathers it all into four commitments that mark the beginning of wisdom.

The chapters are self-contained enough to read out of order — cross-links mark every load-bearing dependency — but the volume was built as one argument, and it compounds. Readers who want the physics behind the branching universe will find it in the volume on the physics of agency; readers who want the formal treatments will find the papers linked where they are used.

This volume is a draft. Chapters carry their status openly, and the arguments are held the way the book says they should be held: at the strength the evidence licenses, and no stronger.