Scottish philosopher Thomas Reid noticed something his contemporaries missed: we don't just form beliefs, we form beliefs about the reliability of our belief-forming processes — and those second-order assessments are themselves vulnerable to the same errors they're meant to catch. This is the core problem with secondary thinking models. When you build a checklist to correct your first-order reasoning, you're still using the same cognitive hardware that produced the original error. The checker and the checked share a blind spot. Reid called our basic inferential capacities 'original principles' — not because they're infallible, but because there's no more fundamental standpoint from which to audit them. The practical implication is discomforting: a second-order model that feels rigorous is not the same as one that is rigorous. The only partial escape is external — a genuinely different person, a genuinely different methodology, or a historical base rate that was generated under conditions entirely unlike your current mental state. Your meta-model needs an outside view more than your object-level model does.

What is the last second-order check you ran that was produced entirely by your own reasoning, with no external data point — and how would you know if it failed?

Drawing from Scottish Common Sense Philosophy — Thomas Reid (Essays on the Intellectual Powers of Man, 1786)