The Objective Pursuit of Knowledge

Why statistical thinking is postmodern.

I was surprised by how divisive last Thursday’s post was. Some thought it was my best post, and others my worst. Let me convince you today that both assertions are true.

I understand why people are confused by the massive uncertainty in health-related decision making and enticed by evidence-based argumentation. When experts disagree, doesn’t that mean that someone is wrong? Shouldn’t there be a rigorous path to find the correct answer? Evidence-based medicine promised a reprieve from ambiguity through cold calculation. Its proponents still insist that biometry alone will lead us to universal truths.

Except, of course, it can’t. As I’ve been yammering on about, medical statistics tell us about population-level optimization and can’t say much about individuals. What a doctor believes about their patient comes from a complex synthesis of expertise and evidence. It never follows simply from a statistical decision-making computer program. EBM both displaces the physician as authoritative and displaces the patient into a statistical aggregate. Moreover, the conditions that establish population-level facts in a trial, like rigid inclusion criteria and strictly controlled experimentation, are simultaneously the conditions that render those facts inapplicable to any individual.

For anyone who’s read a bit of 20th-century French critical theory, that sounds an awful lot like a deconstructive moment. The founders of EBM thought of themselves as working in the pure spirit of Enlightenment rationalism. They were, in fact, unintentionally embracing postmodernism.

Postmodernism, fittingly, means many different things to many different people. Broadly speaking, it’s an embrace of the lack of universal truth, the fluidity of meaning, and the spirit of pluralism. Let me first explain some lessons worth drawing from the postmodernists, who have been far more right in their predictions than prognosticators ought to be. Then I’ll engage with those who receive postmodernism as nihilism and a rejection of plain facts.

Specifically, Jean-François Lyotard articulated how computerization led to an obsession with statistical optimization and an abandonment of grand shared narratives. He declared the postmodern condition to be the replacement of a collective search for universal truth with many disjoint searches for local optimality.

In the postmodern condition, scientific inquiry is reduced to asking whether interventions reduce costs, time, or deaths or increase shareholder value, productivity, or “well-being.” Knowledge is legitimated only if it can do something outside of itself. Research is funded only if it promises to improve performance. Science has to produce things that work. Whether or not any of these produce truth or understanding is irrelevant. Truth is justified through the attainment of ends.

Scaled-up statistical optimization is postmodern, be it in (evidence-based) medicine, economic development, or adtech. Once truth is arbitrated by number-go-up, you disregard explanation, mechanism, and experience in exchange for metric chasing. Lyotard warned us that if knowledge is governed by performance, then centralized technocratic power would get to decide what counts as knowledge and who gets to speak. In 1979, this needed a monograph. In 2026, it’s undeniable.

It is deliciously ironic that mathematical rationality—building statistical models and maximizing expected utility—is the epitome of Lyotard’s postmodernism. Those steeped in rationality would scoff and say that of course statistics isn’t postmodern.¹ Rational calculation was supposed to be the purest form of modern reason. Instead, it has led to a fractured, virtualized, monetized cultural schizophrenia.

This is why the rationalist-adjacent crowd in Silicon Valley has been in a royal tizzy about the postmodernists and their antecedents. They have been fluffing jeremiads against “The Frankfurt School,” decrying postmodernism as the most harmful idea advanced against humanity, all while embracing attitudes toward reality that 20th-century postmodern theorists would have thought caricatures. Geoff Shullenberger has been writing great essays on this topic.²

Now, there’s a problem with pulling out the term postmodern. Lumping a bricolage of related ideas together under a catch-all also made a body of pressingly relevant thought easy to dismiss. I’m sure I’ve already lost people midway through this post because I chose to deploy it. The word “postmodern” tends to end conversations.

Indeed, that’s what happened in the 21st century. Scientists hated postmodernists who questioned their authority. Scientists also romantically put forward the idea that their theories explain universal truths about the fabric of reality and the technology we’ve built on top of it.³ In the 1990s, this led to heated public debates, culminating in the stupid controversy over Sokal’s hoax, in which a theoretical physicist got a fake paper filled with nonsense published in a postmodernism-friendly humanities journal. Postmodernism was thrown aside as unrigorous nonsense.

It was, of course, a pyrrhic victory. In the intervening decades, the academic STEM literature, with LLMs and perverse incentives, has Sokal-hoaxed itself to death. By contrast, the ensuing history has pretty much proven the postmodernists more than right.

So I’m choosing to use the term postmodern because it’s accurate and precise. The term explicitly captures the dialectic of mathematical rationality, the conflation of truth and utility, and the locality of efficacy. And though the French theorists wrote in an annoyingly inscrutable air of pessimism, I’d argue they offer glimpses of alternative political orders. It is perhaps only through productive disagreement that we might break free from the insidious, multiscale pressure of optimization.

1. If you want a hilarious set of hallucinations, try initiating a chat with Claude about whether statistics is postmodern. Its constitution.md file forbids this possibility.

2. Geoff is one of my go-to scholars on this topic. The possibility that evidence-based medicine is postmodern was introduced to me years ago in an episode of a podcast that Geoff hosted during the COVID pandemic. The conversation of this episode influenced my arguments in this and the last few posts.

3. I watched a panel last week that revealed this was still widely held amongst STEM researchers.