When you cannot afford the truth, facts are probability distributions. Nothing is true or false, just likely or unlikely.
The capital-T Truth you learned in school—binary, definitive, certain—is expensive. What we get instead is truth-with-confidence-intervals, facts that exist on a spectrum from "probably" to "almost certainly."
The Cost of Certainty
AI is being honest about how knowledge works. Everything is pattern matching. Your diagnosis isn't true or false—it's 73% likely given symptoms. The historical fact isn't correct or incorrect—it's more or less supported. Additional information comes at a cost, and changes percentages.
Reality is a continuous negotiation with statistical models. You don't get answers. You get probability distributions shaped like answers. The difference matters when the 15% uncertainty contains the black swan that kills you.
People adapt. They speak in probabilities too. "I'm 80% sure the meeting is Tuesday." "High likelihood this works." We're all Bayesians now, updating our priors with each interaction.
We used to have authorities who declared truth, now we have models that suggest likelihoods based on the facts they could access. The priest said "this is true." Statistical AI says "87% probability given available data."
Actual truth—unhedged, binary, certain—requires primary sources, direct observation, mathematical proof. It remains expensive because facts are expensive even when cognitive labor is nearly free.