Axio › Volume 3 › In Defense of IQ

In Defense of IQ

A calibrated score, not a cosmic rank

This chapter is a draft — it is readable but still changing.

There is a mathematically respectable case against IQ, and it deserves an answer on its own terms. It goes like this: cognitive ability spans many dimensions — verbal fluency, spatial reasoning, working memory, processing speed, and whatever else the mind’s machinery varies in. Any single number summarizing performance in that space is a scalar projection: a lossy compression, sensitive to the choice of linear form used to produce it. Project a high-dimensional cloud onto a line and most points land near the middle, so mid-range scores largely reflect geometric concentration rather than substantive differences in intelligence. The measure, on this account, carries little explanatory weight — an artifact of the projection, not a fact about the person.

The mathematical premises are sound. Cognition plausibly does span many dimensions; every summary does compress; compression does lose information. I grant all of it. What I reject is the interpretive step — the leap from “IQ is a lossy projection” to “IQ tells us little.” That inference fails, and seeing why it fails requires asking a question the critique never asks: a projection of what, for what?

Intelligence Is Game-Relative

My answer starts from a commitment developed in the previous chapter: intelligence is effectiveness within a game. A game, in this sense, is any strategic interaction defined by goals, constraints, and choices among alternatives. Strategy presupposes that structure, and intelligence names the effectiveness with which an agent navigates it. Outside a game, intelligence collapses into an undifferentiated abstraction — a word waving at everything and measuring nothing.

Different games reward different competencies. Social coordination rewards signaling and norm navigation. Scientific inquiry rewards hypothesis testing and error correction. Economic competition rewards prediction and optimization under scarcity. Each domain supports its own intelligences, shaped by the strategic environment in which success is evaluated.

This framing carries an immediate implication: every intelligence measure is implicitly indexed to a game. Disputes about intelligence measures arise primarily when that index goes unarticulated — when a score earned in one game is read as a verdict about all of them. So the question about IQ shifts. It is not whether IQ captures intelligence in general — nothing could, because there is no game-free intelligence to capture. The question is whether IQ measures effectiveness within a particular family of strategic contexts, and whether scalar compression is appropriate within that domain.

The Game IQ Measures

IQ measures performance within a recognizable family of games. Their shared structure:

Success in this family depends on rapid pattern extraction, generalization, and error correction under cognitive load. And these competencies matter far beyond the testing room, because much of the modern world is built out of the same structure: educational systems, technical disciplines, bureaucratic institutions — environments organized around formal rules and compressed information flows.

Critics like to note that these institutions resemble the structure of IQ tests themselves, as if that resemblance were a circularity that invalidates the measure. It is not circularity; it is historical path dependence. Games select for the skills that sustain them, and the abstract-symbolic game happens to be the one industrial civilization scaled up. The dominance of that game is contingent — a different history might have organized itself around different competencies — but contingent dominance does not make performance within the game unrankable. Chess ratings are not invalidated by the observation that chess did not have to become popular.

Once the game is specified, the scalar follows naturally. Games generate scores. A score compresses performance into a form that supports comparison, prediction, and coordination under shared rules. Nobody objects that a chess rating fails to capture the player’s kindness or business acumen; its legitimacy depends on whether it tracks success in the game, not on whether it captures every trait of the players. IQ functions the same way. It summarizes effectiveness across many instantiations of the same game structure. Information is lost in compression — that is what compression means — yet relevance is preserved so long as the summary retains the variance that predicts future performance within the domain.

Anchored, Not Arbitrary

Here is where the projection critique’s strongest claim — that the scalar is sensitive to an arbitrary choice of linear form — meets its empirical answer. IQ was not conjured by picking a direction in cognitive space and projecting onto it. It emerged from an observation: performance across many abstract cognitive tasks covaries. People who do well on one tend to do well on the others. That shared variance — the general factor, \(g\) — is a discovered pattern, not a chosen axis.

Test construction follows the signal. Tasks that load strongly on the shared variance are retained; tasks that do not are discarded. The process is self-constraining: variants that drift away from the underlying structure lose predictive power and get culled. The measure’s stability across test batteries, populations, and decades reflects anchoring to a real pattern, not freedom of representation. A genuinely arbitrary projection would not survive that filter. Within the game it measures, IQ behaves like a calibrated score, not an arbitrary one.

The bell-shaped distribution earns a similar reading. Human cognitive performance develops within biological and environmental limits — neural architecture, developmental pathways, energetic constraints bound the variance long before any test is administered — and aggregation across bounded traits produces familiar distributions. Biology bounds the space of possible cognitive strategies; games determine which regions of that space are amplified, rewarded, or ignored. The curve reflects the structure of the underlying system, not an artifact of projection geometry alone, and it persists across contexts because the game’s constraints are stable.

What about generality? The critic can concede all this and still insist the score is parochial. But general intelligence never required universality across all games. It requires transfer across a broad class of games sharing deep structural features — and abstract reasoning, symbolic manipulation, and error correction recur across many strategic environments, which is why competence in one supports competence in others. IQ tracks effectiveness across this equivalence class. That explains both its transferability and its limits in one stroke: it is neither a game-free essence nor a narrow parochial skill, but a measure of performance across a structurally related family of games.

And within that family, the predictive record is not seriously in dispute. IQ correlates with learning rate, symbolic reasoning, educational attainment, performance in cognitively dense institutions — even health and longevity — and these correlations are among the strongest and most reproducible in social science. They operate probabilistically and admit wide individual variation, but predictive reach is what justifies a measurement. What it does not do is interpret itself. Interpretation requires premises from outside the game.

When Scores Escape Their Game

So where does the controversy come from? Almost entirely from scores escaping their game.

One standing critique — Taleb’s is the best known — charges that IQ fails to measure creativity, emotional intelligence, practical wisdom, resilience, adaptability in complex and unpredictable environments. Every item on the list is true, and none of it is an objection. It is a list of other games. A measure indexed to the abstract-symbolic family owes us no verdict on domains outside that family, and holding the absence of one against it is like faulting a chess rating for not predicting poker winnings. The critique lands only against people who wield IQ as a universal rank — and against them it lands squarely. The measure is innocent; the misuse is real.

A second confusion is political. Scalar scores influence coordination and status, and that influence generates discomfort — discomfort that gets misdirected at the measurement rather than at the uses it is put to. Why do modern societies reward this particular game so lavishly? A fair question, and one about political economy, not about measurement validity. Conflating the two obscures both.

The third confusion is the deepest, and it explains the peculiar heat of the whole debate: the conflation of cognitive capacity with moral worth. Much of the vehement rejection of IQ differences, particularly from egalitarian quarters, rests on a suppressed premise — that if people differ measurably in cognitive ability, they must differ in moral standing. Accept that premise and denying the differences becomes a moral emergency. But the premise is exactly wrong, and the value framework of this book says why: there is no objective scale of value on which a game score could register as cosmic rank. Moral worth and dignity attach to personhood, not to percentile. Recognizing empirical differences in cognitive abilities implies nothing about moral inequality, justifies no discriminatory treatment, diminishes no one’s dignity — any more than differences in chess ratings do.

Getting this distinction clear is not a concession to the critics; it is what disarms them. Once intelligence and moral value are cleanly separated, the defensiveness becomes unnecessary and honest conversation becomes possible — about education, about policy, about tailoring interventions to actual rather than pretended distributions of ability. Denying empirical realities out of moral discomfort does not protect anyone. It just guarantees that the interventions miss.

What the Score Doesn’t Buy

If IQ is a calibrated score within a real game, one last question remains: what does an extreme score actually purchase? Consider three people who would saturate any plausible test — Terence Tao, Stephen Wolfram, David Deutsch — and notice that in each case the score is the least interesting fact about them.

Tao’s career is a study in depth. A child prodigy who never stopped, he committed himself to mathematics with a relentlessness that produced breakthroughs across analytic number theory, combinatorics, and partial differential equations — discoveries unreachable by any broad or superficial approach. Wolfram’s is a study in translation: he did not stop at theoretical work on computational complexity but built Mathematica and Wolfram|Alpha, converting abstract insight into tools used by millions. Deutsch’s is a study in synthesis: quantum computation and the Many-Worlds Interpretation, merged with epistemology and evolution into a philosophical framework ambitious enough to reshape how physics understands itself.

Three different games, note — depth, translation, synthesis — all drawing on the same abstract-symbolic core that IQ measures. That is the transfer claim made flesh. But look at what did the compounding. All three paired talent with sustained discipline: their achievements were not episodic but the yield of decades of consistent production. All three showed intellectual courage — Wolfram arguing that complexity emerges from simple computational rules, Deutsch defending an interpretation of quantum mechanics that drew years of skepticism, Tao walking into conjectures reputed to be impossible. All three followed intrinsic curiosity rather than any predefined trajectory; the breakthroughs emerged from the following, not from a plan. And the two who learned to communicate — Wolfram and Deutsch through books and public argument — multiplied their influence far beyond their fields.

Discipline, courage, curiosity, communication: none of these appears on an IQ test, and the greatness is unaccountable without them. Which is the defense of IQ in miniature. The score is real — anchored to shared variance, predictive within its family of games, as legitimate as any calibrated measure of performance we possess. And it is a score, not a destiny: a measure of position within one game, silent about the others, and silent above all about what a player will build from the position. Defending IQ in this sense is defending disciplined abstraction — bounded by context, anchored in structure, and evaluated by what it enables rather than by what it excludes.