Matters of Concern

In the AI Story

The distinction emerged from Latour's late-career engagement with critiques of science — particularly the criticisms he himself had been accused of enabling. In Why Has Critique Run Out of Steam? (2004), Latour worried that the critical deconstruction of scientific facts, which he had helped develop for analysis of well-established and stable science, had been weaponized by climate-change deniers and other bad-faith actors to reopen settled questions. His response was not to abandon critique but to refine it: the goal was not to dissolve all facts into social construction but to distinguish facts that deserved their status from those that circulated as facts when they were actually matters of concern pretending to be settled.

The framework reads as though written for the AI discourse. The claim that AI produces twenty-fold productivity gains is sometimes a report from a specific, heterogeneous network — experienced engineers, a particular AI tool, a specific organizational culture, a particular kind of feature being built. In that concrete context it is a concern, contestable, entangled with the network that produced it. In its travels — into slide decks, board presentations, media reports — it sheds its qualifications and arrives at its destination as a fact. The network origins are purified away. The number is now universal, applicable to any context, and actionable without deliberation.

The same operation applies to the democratization thesis. The claim that AI democratizes capability is real in the narrow sense that the model is deployed uniformly. But democratization is not a fact; it is a concern. The model requires connectivity, hardware, English fluency, and cultural capital to use effectively. The developer in Lagos accesses the same model as the engineer at Google but operates in a different network with different institutional support. To treat democratization as a fact is to foreground the uniform deployment and background the asymmetric conversion factors that determine what the deployment actually produces. The matter of concern — under what conditions does AI deployment translate into substantive capability expansion? — is closed before it can be deliberated.

The conversion of concerns into facts is not a conspiracy. It is a structural tendency of discourse under conditions of speed. Clean numbers travel faster than qualified analyses. Settled claims are easier to act on than contested ones. The incentive structures of media, government, industry, and academia all reward fact-production and penalize concern-maintenance. The result is a discourse in which the most important questions — who benefits, who bears costs, whose perspectives shape the training data — are buried under a surface of confident numbers and settled claims.

Origin

The distinction was developed most influentially in Latour's 2004 Critical Inquiry essay Why Has Critique Run Out of Steam?, though its roots extend through his entire career of laboratory ethnography. The immediate provocation was the use of science-studies vocabulary by climate-change deniers — an alliance Latour found both intellectually humiliating and politically urgent to address.

The reframing was also an invitation: rather than asking whether claims are 'really true' or 'socially constructed' — a question the fact/construct binary could not usefully answer — ask instead whether a claim is being treated as a matter of fact (stable, uncontested) or a matter of concern (open for renegotiation), and whether that treatment is appropriate given the claim's actual epistemic status.

Key Ideas

Facts as a status, not a property. A claim becomes a fact through the work of stabilization. Treating a concern as a fact is not an innocent shortcut; it is a political operation that forecloses deliberation.

Purification as rhetorical strategy. Converting a concern into a fact typically involves stripping the network origins away — presenting the residue as universal when it is actually context-specific.

The speed bias of discourse. Institutions reward clean claims that travel. Matters of concern, which require qualification and deliberation, are structurally disadvantaged in the attention economy.

Re-opening is the critical move. The work of critique in the AI age is not to produce more facts but to convert facts back into concerns — to re-introduce the political, evaluative, and distributional dimensions that purification strips away.

Governance prerequisite. Matters of concern require deliberation; matters of fact require only verification. Governance that treats concerns as facts short-circuits the deliberative process that gives democratic legitimacy to technological transformation.

Debates & Critiques

Defenders of purification argue that pragmatic decision-making requires settled claims: you cannot deliberate every question forever, and at some point the network must be black-boxed so that action becomes possible. Latour's reply accepts the pragmatic point while insisting on the distinction between appropriate and premature closure. A claim like 'water boils at 100°C' has been stress-tested through centuries of experimental refinement; treating it as a fact is justified. A claim like 'AI produces twenty-fold productivity gains' has not; treating it as a fact is premature. The question is not whether to close questions but whether the closure has earned its status through the kind of sustained, multi-stakeholder testing that legitimate factuality requires.