Through Kuhn’s Lens

Regulation Outpaces Skills Vocabulary

September 07, 2026 | 2529 words

Through Kuhn’s Lens: A Discourse That Can Only Speak of Rules

Count the words. That is where this week’s phenomenon begins, and it is a matter of counting, not impression. AI governance discourse has an enormous vocabulary for regulation—compliance, oversight, risk tiers, enforcement, audits, penalties, guardrails, mandates. It has a starved vocabulary for what happens after the rules are written: how people learn to work with the technology, how they build partnerships with it, how their labor transforms in contact with it. The imbalance is not a mood. It is a fact about which words the discourse stocks and which it lacks.

Notice what this does. Society is being asked to govern a technology using a language built almost entirely for governing—and almost not at all for using. The discourse can specify what may not be done. It struggles to say what working with the technology even looks like. When a vocabulary lacks words for “skills,” “partnership,” and “transformation,” it is not merely missing adjectives. It is missing the capacity to see certain things at all. That is the phenomenon on the table, and Kuhn’s framework offers a precise instrument for reading it.

Normal Science on Governance

Thomas Kuhn’s central distinction, laid out in The Structure of Scientific Revolutions, separates two modes of intellectual work. Normal science is puzzle-solving inside an accepted frame—a paradigm—where the community agrees on what counts as a problem, what counts as a solution, and which tools apply. Revolutionary science is what happens when that frame breaks: when the accepted problems stop being solvable and the community must reframe what a problem even is.

Read through this lens, the regulation-heavy discourse looks like normal science. It is not confused or empty. It is productive. It solves puzzles briskly and well. Given a new model release, it asks: which risk tier, which disclosure requirement, which liability regime? These are real puzzles, and the discourse has real tools for them. The vocabulary of regulation is not thin because the community is lazy. It is thick precisely because the community has been solving governance puzzles inside a stable frame for years.

This is the first thing Kuhn’s framework clarifies, and it cuts against a lazy reading. The imbalance is not a failure of effort. It is the signature of a paradigm working normally. A mature normal-science tradition develops a dense, specialized vocabulary for the problems it recognizes—and a corresponding silence about problems it does not. The density of regulatory language and the poverty of skills language are the same fact seen from two sides. The frame that makes one rich makes the other invisible.

So the diagnostic question is not “why is the discourse unbalanced?” It is: what is the frame that produces this particular balance, and what accumulates underneath it that the frame cannot name?

The Anomaly a Regulatory Vocabulary Cannot Register

Kuhn’s term for the thing a paradigm cannot name is anomaly—an observation the accepted frame did not predict and cannot comfortably absorb. In The Structure of Scientific Revolutions, anomalies are the pressure that eventually cracks a paradigm. But here is the subtlety the case demands: an anomaly must first be seen to exert pressure. A frame that lacks the vocabulary to register a phenomenon does not experience it as an anomaly. It experiences nothing at all.

This is the heart of the matter. A vocabulary built for regulation registers a specific kind of event: a harm, a violation, a risk to be bounded. It is exquisitely tuned to catch what goes wrong. It is deaf to what goes right—to the quiet accumulation of people actually working with the technology, forming partnerships with it, being transformed by it in ways that are neither harm nor compliance.

Consider what “partnership” would name if the discourse had the word. It would name the space between the human and the system where work actually happens—where a person delegates, corrects, negotiates, and adapts. That space is where nearly all the real consequences of this technology are being produced right now. And the governance vocabulary has almost no words for it. The discourse can tell you whether a model is permitted. It cannot tell you whether a partnership with it is good.

This is the anomaly the regulatory vocabulary papers over: the ordinary, non-catastrophic transformation of work is happening at scale, and the dominant frame has no instrument to detect it. Not because the transformation is hidden, but because the measuring-stick only reads danger. A thermometer cannot report color. A vocabulary of rules cannot report a partnership.

The danger here is not that the anomaly is suppressed. It is subtler and worse. The anomaly is not experienced as an anomaly, because the frame that would feel its pressure has no receptor for it. Under normal science, this is precisely how a paradigm defers its own crisis. It keeps solving the puzzles it can name and keeps not-seeing the phenomenon it cannot.

Two Communities, Two Measuring-Sticks

Kuhn’s most useful concept for this case is incommensurability. In plain English: two communities can use different measuring-sticks, so that a claim decisive in one is not merely disputed in the other—it is unintelligible there. It does not translate. It fails to register as a claim at all. Kuhn spent his last years refining this idea; the fullest statement is in The Last Writings — Incommensurability in Science, where he locates incommensurability not in disagreement but in vocabulary—in the specific words each community has and lacks.

That is exactly the fault line this week’s phenomenon reveals. There are two discourse communities here. One is governance-and-regulation. The other is skills-and-practice—the people actually building working relationships with these systems. And they do not disagree so much as fail to reach each other.

Kuhn’s term for a community’s standard case of “what a solved problem looks like” is exemplar—the shared model that trains members to see problems the same way. Compare the exemplars.

For the governance community, the exemplar of a solved problem is a well-drafted rule. A clean risk classification. An enforceable disclosure requirement. A liability regime that assigns responsibility when something breaks. Success looks like a boundary correctly drawn.

For the skills-and-practice community, the exemplar of a solved problem is a working partnership. A person who has learned where the system is reliable and where it is not. A team that has restructured its labor around the technology’s actual grain. Success looks like a capability gained, not a boundary drawn.

Now watch the incommensurability. Present the governance community with a beautiful account of a transformed workflow, and it asks: but is it compliant? The transformation does not register as an achievement—only as a site of unaddressed risk. Present the practice community with a flawless risk taxonomy, and it asks: but does it help anyone actually work? The taxonomy does not register as help—only as a constraint on work already underway. Each community’s decisive win is the other’s non-event.

This is not stubbornness. It is structural. The measuring-sticks are different, so the achievements do not convert. A partnership cannot be scored on a compliance scale, and a rule cannot be scored on a capability scale. And because the governance vocabulary is the one that dominates public discourse, its measuring-stick becomes the default—the one society reaches for when it tries to think about AI at all. The imbalance is not two communities holding equal ground. It is one measuring-stick crowding out the vocabulary that would let us read the other.

The statistics this week make the crowding measurable. Where the data shows governance discourse dominated by regulatory terms and starved of skills, partnership, and transformation language, it is documenting the relative supply of two vocabularies competing to frame the same technology. That imbalance is the incommensurability made countable. It tells us not that one community is wrong, but that one measuring-stick has become the public default while the other has almost no words in circulation. When a society can generate a dense taxonomy of things a technology must not do, and can barely form a sentence about what working with it well would mean, the vocabulary gap is doing real damage—it is shaping what the public can even think to ask.

Policing the Term

Someone will call this a paradigm shift. Vendors especially. The phrase is loose currency now, spent on every product cycle. A Kuhn column must police it, because Kuhn’s machinery is far more demanding than the colloquial use.

In The Structure of Scientific Revolutions, a paradigm shift is not a big change or an important moment. It is a specific sequence: a stable paradigm, an accumulation of anomalies it cannot absorb, a period of crisis in which the community loses confidence in the frame itself, and then a revolution in which a new frame reorganizes what counts as a problem. Each stage is a threshold. Skip one and you do not have a shift. You have change.

Apply the sequence to the vocabulary imbalance. Is there a stable paradigm? Yes—the regulatory frame governs the discourse. Are anomalies accumulating? Here the case grows precise and uncomfortable. The transformation of work is happening, but the frame does not register it as anomalous. So the pressure that would drive a crisis is not building inside the discourse. It may be building in the world. It is not building in the words.

That distinction is decisive, and it is where the colloquial “paradigm shift” collapses. Kuhn’s crisis is not an external condition. It is an internal one—a community’s felt loss of confidence in its own frame. There is little evidence of that here. The governance community is not in crisis. It is confident, productive, expanding its vocabulary in the direction it already faces. That is not a paradigm on the verge of revolution. That is a paradigm doing exactly what a healthy paradigm does: solving its puzzles and not noticing what it cannot name.

So the correct Kuhnian term for this moment is not “crisis” and not “revolution.” It is closer to what Kuhn called a pre-paradigm condition for the missing vocabulary—the skills-and-practice frame has not yet cohered into a shared paradigm with its own exemplars and terms. The regulatory paradigm is mature. The partnership paradigm barely exists. What we are watching is not one frame breaking, but a second frame failing to be born.

This is a more austere diagnosis than “paradigm shift,” and a more useful one. It tells us the problem is not that the old frame is cracking. The problem is that the vocabulary needed to see the anomaly does not yet exist. And a paradigm cannot be broken by an anomaly it has no words to feel.

What a Real Shift Would Look Like

Be demanding here. A paradigm shift is not “more balanced language.” Balance is a cosmetic target—one could stuff the discourse with the word “partnership” and change nothing. Kuhn’s criterion is stricter. A shift would show up as a change in what counts as a solved problem.

In The Copernican Revolution, Kuhn traced a real one. The shift from an Earth-centered to a Sun-centered cosmos was not the arrival of new words. It was a reorganization of what an astronomer’s problems were, which observations mattered, and what an explanation had to do. After the revolution, the same sky posed different puzzles. The exemplars changed.

So the test here is not vocabulary volume. It is this: has the governance community’s exemplar of a solved problem changed? A genuine shift would look like a regulatory body treating a well-formed partnership as a first-class object of governance—something to be assessed, cultivated, and reasoned about, not merely bounded. It would look like “transformation” entering the discourse not as a vendor’s boast but as a governable phenomenon, with its own criteria for going well or badly. It would look like the compliance measuring-stick and the capability measuring-stick becoming commensurable—translatable into each other—so that a claim about skills could count as evidence in a governance argument.

That has not happened. And Kuhn’s refinements in The Essential Tension, especially his second thoughts on what a paradigm actually is, warn against mistaking the appearance for the substance. A community can adopt new words while keeping its old exemplars. It can say “partnership” while still scoring everything on the compliance scale. The words drift; the measuring-stick holds. That is not a revolution. That is the old paradigm assimilating new vocabulary into unchanged practice—normal science absorbing a term, not a shift dislodging a frame.

A real shift would be recognizable by friction, not fluency. The community would find its old puzzles suddenly harder to state. Its confident risk taxonomies would start feeling like they were answering the wrong question. That discomfort—the felt inadequacy of the frame—is the mark of an approaching crisis. Its absence is the mark of its absence.

The Reader’s Stake

Strip the machinery and say plainly what is at issue for the reader, because this is where power operates.

A discourse that can only speak of regulation hands the public a single way to relate to this technology: as something to be managed, feared, and bounded. That framing is not neutral. It serves the institutions whose authority is regulatory. When the only available vocabulary is the vocabulary of rules, the only available public role is that of the governed—a spectator to a technology others will permit or forbid on your behalf.

The missing words are the words that would return agency to the reader. “Skills” names something a person can acquire. “Partnership” names something a person can form. “Transformation” names something that happens to and through the person, not merely around them. A vocabulary that lacks these terms does not just fail to describe the technology. It fails to describe the reader’s own possible relationship to it. It leaves the public able to ask whether AI should be allowed, and unable to ask what it would mean to work with it well.

That is the anti-mystification point, and it is the sharp one. The vocabulary imbalance is not only an intellectual gap. It is a distribution of agency. And Kuhn’s framework, applied honestly, refuses to let the imbalance hide as a mere matter of emphasis. It is a matter of which questions the public is equipped to ask—and therefore of who gets to answer them.

The Falsification Condition

A Kuhn column ends by naming the evidence that would move its own reading. So here is the question this analysis cannot yet answer: is the missing vocabulary a symptom of an approaching crisis—or simply a discourse that has not gotten around to naming its tools yet?

Those are genuinely different states, and the current data cannot distinguish them. Both look identical from the outside: a discourse rich in regulation, poor in partnership. The difference is internal—whether the governance community has begun to feel its frame as inadequate.

State the falsification condition plainly. Watch the governance community’s exemplars, not its vocabulary. If, over the coming period, regulatory bodies begin treating well-formed partnerships as governable objects—if “transformation” enters governance discourse as a phenomenon to be assessed rather than a slogan to be dismissed, if compliance arguments start admitting skills-evidence as decisive rather than irrelevant