The Machine That Trusted Itself

Nicholas Rescher published this book in 2003 with the calm confidence of a man who had been thinking about epistemology for half a century and saw no reason to rush. The result is a textbook that reads like a philosophical testament — systematic, thorough, and almost defiantly pre-digital in its concerns. Rescher's central architecture is pragmatic coherentism: knowledge is not built on unshakeable foundations but assembled from plausible truth-candidates, tested against each other, validated by their practical efficacy. Fallibilism is the governing mood. We are always wrong about something; the point is to be wrong in ways we can correct. In 2003, this felt like a mature, stabilizing position — the epistemological equivalent of a well-diversified portfolio. In 2026, it reads like a warning letter that arrived twenty years early and was filed without being opened.

What Rescher got right, with an almost eerie precision, is the centrality of trust and cooperation in the economy of knowledge. Chapter 6's analogy between epistemic trust and financial credit now looks less like a pedagogical metaphor and more like a structural diagnosis. He argued that trustworthiness is built incrementally, that monopolizing information is counterproductive, and that the rational default is cooperative sharing. He was describing, without knowing it, the architecture that would underpin Wikipedia, open-source science, and the early promise of social media — and also the architecture that would collapse when those systems encountered adversarial actors, algorithmic amplification, and state-sponsored disinformation at scale. His framework anticipated the stakes but not the weapons. The idea that rational agents in communities naturally evolve toward "normatively cogent reasons" through a process of rational selection (Chapter 10) now reads as the most optimistic sentence in the book. We have watched communities of agents — some human, some not — evolve toward engagement maximization instead, and the selection pressure was anything but rational.

The deepest blind spot is not technological but ontological. Rescher spends considerable energy on the question of whether alien intelligences might practice science radically different from ours (Chapter 16), a thought experiment that was charmingly speculative in 2003. He did not anticipate that the more pressing version of this question would involve intelligences we built ourselves. Large language models do not practice science, but they do produce text that is coherent, plausible, and often wrong in ways that are structurally invisible to casual inspection — which is to say, they are Gettier problems at industrial scale. Rescher's careful distinction between weak and strong epistemic justification, his insistence that experience alone provides only weak justification for objective claims, his entire apparatus for thinking about presumption as a "tentative gap-filler" — all of this becomes urgently relevant when the gap-filler is a probabilistic model generating answers with no grounding in truth-commitment whatsoever. The book has the tools. It simply could not have imagined what would need them.

Within the broader epistemological tradition, Rescher occupies a specific and somewhat lonely position: too pragmatic for the analytic purists, too systematic for the pragmatists, too continental in his scope for the Anglo-American mainstream. He draws heavily on Peirce's fallibilism, nods to Quine's web of belief, and builds outward from the coherentist tradition while refusing to abandon realism entirely. The result is a book that serves as a bridge — between the Gettier debates of the 1960s and the social epistemology that would flourish in the 2010s, between classical philosophy of science and the science-and-technology studies that were already reshaping the field when he wrote. His Law of Logarithmic Returns (Chapter 13), which holds that significant scientific findings become harder and costlier to extract as knowledge accumulates, has been quietly vindicated by two decades of replication crises, ballooning research budgets, and diminishing returns in pharmaceutical discovery. He was not the first to notice this pattern, but he gave it a name and a formal shape that has aged well.

What strikes hardest now is Chapter 3's dismissal of skepticism as "economically and practically irrational." Rescher meant that demanding absolute certainty before acting is a luxury no functioning agent can afford. He was right. But the inverse has also proven true: accepting claims without adequate justification because the cost of skepticism feels too high is its own catastrophe, and we have spent the last decade living inside it. The book assumes a world in which the primary epistemic danger is excessive doubt. We live in one where the primary danger is excessive credulity — not because people are foolish, but because the information environment has been engineered to make credulity the path of least resistance. So the question the book now raises, which it could not have raised in 2003: If fallibilism tells us we are always wrong about something, and if the systems we have built are designed to make it harder to discover *what* we are wrong about, does pragmatic coherentism still function — or does it simply ratify whichever errors are most coherent with each other?