Tests of Worth
Tests of worth are the concrete evaluative moments through which each order of worth operationalizes its grammar of value. In the inspired order, the test is the capacity for genuine creative expression — a painting, a theorem, a performance that carries the mark of originality. In the market order, the test is the transaction — goods or services exchanged at a mutually agreed price. In the industrial order, the test is efficiency — output per unit of input, measurable and comparable. Each order's tests give it institutional reality: without tests, the order is a rhetoric rather than a practice. AI disrupts tests across multiple orders simultaneously, producing the distinctive confusion of the current moment.
In the AI Story
The concept is central to Boltanski and Thévenot's framework because it connects abstract grammars of justification to concrete practice. Disputes are not resolved by appeal to philosophical principles; they are resolved by tests that parties accept as authoritative. The worker and the manager do not argue about the meaning of productivity; they argue about what counts as a valid measurement of it. The artist and the critic do not argue about the meaning of originality; they argue about whether this specific work demonstrates it.
AI creates a distinctive category of disruption because it passes the surface of multiple tests without satisfying their underlying purposes. AI-generated text passes plagiarism checkers, grammar tests, and coherence evaluations — yet the question of whether it constitutes authentic creative expression remains unresolved. AI-generated code passes compilation and runs without error — yet the question of whether it represents genuine engineering judgment remains open. The tests were designed when only humans produced the outputs; they measured features correlated with human practice rather than features constitutive of it.
The result is what Boltanski's framework would call a crisis of tests. The evaluative infrastructure on which professional legitimacy depended is being systematically outmatched by systems that can produce test-passing output without the underlying capacity the test was meant to detect. This is distinct from the familiar problem of gaming metrics. The AI is not gaming; it is producing genuine output. The problem is that the tests were measuring a proxy for what mattered, and the proxy has detached from what it proxied.
The response cannot be to abandon tests — evaluation is necessary for any functioning practice. The response must be to redesign tests so they measure what they were originally meant to measure. This requires understanding what cannot be faked by AI: embodied judgment, situated practice, the accumulated relational knowledge that Ericsson's research on deliberate practice documented.
Origin
Boltanski and Thévenot introduced tests of worth in On Justification as the practical infrastructure of each grammar of evaluation. The concept connects their framework to the pragmatic tradition in sociology — focused on how people actually do evaluation, not how they abstractly theorize it.
Key Ideas
Concrete evaluation. Tests are the practical moments where grammars of worth operate in specific disputes.
Proxy detachment. AI produces outputs that pass tests designed as proxies for human capacity without instantiating the capacity.
Crisis across orders. AI disrupts tests in inspired, industrial, market, and projective orders simultaneously.
Redesign imperative. Tests must be redesigned to measure what cannot be faked: embodied judgment, situated practice, relational knowledge.
Beyond gaming. The AI test-passing problem is not metric-gaming; it is proxy-detachment.
Appears in the Orange Pill Cycle
Further reading
- Luc Boltanski and Laurent Thévenot, On Justification: Economies of Worth (Princeton, 2006)
- Laurent Thévenot, L'action au pluriel (La Découverte, 2006)
- Luc Boltanski, On Critique: A Sociology of Emancipation (Polity, 2011)