In Defense of IQ
Summary
This essay responds to critiques framing IQ as a lossy scalar projection from high-dimensional cognitive space by grounding the discussion in a prior commitment: intelligence is game-relative effectiveness. Intelligence only becomes meaningful relative to strategic contexts defined by goals, constraints, and choices. IQ measures performance within a specific family of games characterized by abstract rule systems, symbolic manipulation, time-bounded decisions, and formal transfer. Within this domain, scalar measurement is appropriate because games generate scores—compression that tracks success within the game. The essay argues IQ’s legitimacy derives from empirical constraint (performance covariance across tasks), stability, and predictive reach within its domain. General intelligence (g) requires not universality across all games but transfer across structurally related families. The piece separates measurement validity (does IQ track effectiveness in its domain?) from political questions (why do societies reward this game?), defending IQ as disciplined abstraction bounded by context.
Key Concepts
- Intelligence is game-relative – Strategy presupposes goals, constraints, alternatives; intelligence = effectiveness navigating that structure; no coherent definition outside specific games.
- The game IQ measures – Family of games sharing: abstract rules, symbolic manipulation, time-bounded decisions, formal transfer; rewards pattern extraction, generalization, error correction under load.
- Scalar legitimacy within games – Games generate scores; compression valid when it retains variance predicting future performance; controversy arises from treating scores as universal.
- Empirical constraint – IQ emerged from observed covariance across tasks; test construction retains high-loading items; stability reflects anchoring to real pattern, not arbitrary choice.
- Bounded variation – Neural architecture, development, energetic constraints bound cognitive space; bell curves reflect system structure, not just projection geometry.
- General intelligence as transfer – g doesn’t require universality; requires effectiveness across structurally related game families (abstract reasoning, symbolic manipulation, error correction).
- Predictive reach – Within domain, IQ associates with learning rate, symbolic reasoning, institutional performance; probabilistic, admits individual variation.
- Measurement vs. valuation – IQ describes effectiveness in defined context; political/social decisions require additional criteria; why societies reward this game = political economy question.
Evolution Notes
- Applies the “intelligence as game” framework (established in earlier post) to defend IQ measurement against dimensionality critiques.
- Shifts debate from “does IQ capture all intelligence?” (obviously not) to “does it validly measure effectiveness in its domain?” (yes).
- The game-relative framing dissolves many traditional IQ debates by clarifying scope.
- Distinguishes measurement validity from status/coordination consequences—separates technical from political questions.
- Defends scalar compression as natural consequence of game structure rather than reductive error.
- Positions IQ critique as often category error: attacking measurement for consequences of how societies use it.
Tags
Cross-References
Open Questions
- If IQ measures effectiveness in games shaped by modern institutions, does circular reinforcement eventually decouple it from “natural” cognitive capacity?
- How should societies weight effectiveness in IQ-like games vs. other intelligence domains (social, creative, practical) when allocating resources/status?
- Can the game-relative framework identify new intelligence measures for domains currently under-measured?
- Does the stability of IQ across populations/time indicate deep biological constraint, or convergent institutional selection?
- If games select for skills that sustain them, what evolutionary pressures shaped pre-institutional human intelligence?
- How does AI performance on IQ-like tasks (pattern matching, symbolic manipulation) relate to this framework—same game or different?