Article
Probability and fine-tuning
Constants, coincidence, and what Bayesian language does — and does not — license about design.
Fine-tuning arguments live or die on careful probability talk. If the constants of nature appear delicately balanced for complexity, that is a fact about parameter space worth stating clearly. What you may infer from that fact depends on your prior commitments: about the measure over possible laws, about selection effects, about whether a designer is a live hypothesis. Hand-waving “it just looks designed” and hand-waving “anthropic principle, end of story” are equally unserious.
Bayes’ theorem does not care about your politics. Likelihoods must be specified; priors must be defended; alternative hypotheses must be stated so that they can be compared. Multiverse proposals that multiply unobservable domains to soak up improbability are not free — they are models with their own costs. Design hypotheses are not free either — they require content beyond “something did it.” Intellectual honesty means pricing both.
The scientific question is how constrained the observed constants are under a given measure. The philosophical question is what that constraint means for theism. Those questions interact, but they are not identical. Treating a order-of-magnitude estimate as a knockdown proof — for either side — is rhetoric wearing a lab coat.
More to come: worked examples of where fine-tuning claims are strong, where they are soft, and how to keep probability language from becoming a confidence trick.