Kaizen Grey

Facts Were Never Going to Be Enough

There’s a strange pairing sitting in my head lately: Thomas Hobbes, writing in 1651 about why a sovereign has to control the meaning of words, and Newt Gingrich, writing in 1990 in a memo that reads like a campaign manual for the same idea. Three centuries apart, one trying to prevent civil war and one trying to win a House majority — and they land on the exact same insight. Language isn’t a neutral container you pour truth into. It’s a mechanism of control, and whoever controls it wins the argument before the facts even show up.

Liberals, broadly, don’t believe that. Or don’t want to. And it’s costing us.

What Hobbes actually said

Hobbes’s Leviathan isn’t usually read as a media theory text, but Chapter 4, “Of Speech,” basically is one. He lays out what he calls the “abuses of speech” — the ways people use words to register their thoughts inconsistently, to speak metaphorically in order to deceive, and to make others believe things are true that aren’t. His famous line: words are “wise men’s counters, they do but reckon by them; but they are the money of fools,” who take a word’s face value as if it settled the argument on its own [1].

But the sharper point comes later, in Chapter 29, where Hobbes diagnoses what actually dissolves a commonwealth. High on the list: private citizens deciding for themselves what words like “tyrant” mean, and using that private definition to justify rebellion [1]. His solution has nothing to do with letting the best argument win. It requires the sovereign to be the final arbiter of what words and doctrines mean, because a society where everyone gets to define “justice” or “tyranny” for themselves doesn’t stay a society for long. For Hobbes, controlling the vocabulary is statecraft, not an abuse of power — the precondition for peace itself.

What Gingrich actually did

Fast-forward to 1990. Gingrich, then a minority-party congressman in a House Republicans hadn’t controlled since 1955, distributed a memo through GOPAC titled, without irony, “Language: A Key Mechanism of Control.” It’s a literal glossary — 133 words split into two columns. “Optimistic Positive Governing Words” to use about yourself and your allies: common sense, freedom, courage, prosperity, moral, principled, dream, truth. “Contrasting Words” to use about opponents: decay, failure, shallow, sick, pathetic, corruption, radical, traitors, hypocrisy, welfare. The cover letter told candidates this wasn’t guesswork — the words were “tested language from a recent series of focus groups where we actually tested ideas and language” [2]. Candidates were told to use this vocabulary in speeches, mailers, and interviews, not to describe reality more precisely, but to make reality feel a certain way before anyone checked.

It worked. Four years later Republicans took the House for the first time in forty years, and the “Contract with America” campaign that got them there is remembered as much for its language discipline as its policy content. Gingrich didn’t invent the idea Hobbes described — he operationalized it as a repeatable campaign tactic, with the receipts to prove it worked.

The line from Gingrich to now

This isn’t a one-off. Frank Luntz, the GOP’s most influential pollster, built a career testing which words move people and handing the winning ones to Republican officeholders. His 2002 memo on environmental messaging is the clearest case: he told Republicans “the phrase ‘global warming’ should be abandoned in favour of ‘climate change,’” and conceded the party had “lost the environmental communications battle” on the merits, so the fight had to move to vocabulary instead [3]. Estate tax became “death tax.” Tax cuts became “tax relief,” a word that presupposes taxation is an affliction before anyone’s argued the point. (For what it’s worth, Luntz later disavowed this specific chapter of his career, telling a Senate committee in 2019, “I was wrong in 2001. Just stop using something that I wrote eighteen years ago, because it’s not accurate today” [4] — which I’d rather note than leave out, since the point here is the mechanism, not that everyone who’s ever used it is unrepentant.)

That’s Hobbes’s insight wearing a focus-group badge: control the word, and you’ve already won the frame the fact has to live inside.

Why liberals keep losing this specific game

Here’s the part that isn’t a hot take — political scientists Matt Grossmann and David Hopkins made this case rigorously in Asymmetric Politics [5]: the Republican Party functions as an ideological movement organized around a coherent worldview and vocabulary, while the Democratic Party functions as a coalition of interest groups, each pursuing its own specific, accurate, defensible piece of the truth. Labor wants its language. Environmental groups want theirs. Civil rights organizations want theirs. Each faction is right about its own corner, and each insists on saying so precisely — which means the coalition as a whole never converges on one repeatable sentence the way “tax relief” or “death tax” did.

George Lakoff made the linguistic version of the same argument in Don’t Think of an Elephant [6]: Republicans think in frames, Democrats think in facts, and a fact recited inside your opponent’s frame reinforces the frame, not the fact. Arguing against “tax relief” by explaining why the tax in question is fair still uses the word “relief” — which has already told the listener taxation is a burden to be relieved of. You can win the sentence and lose the frame.

Liberals treat word-choice-for-effect as beneath us, as if it’s itself a kind of dishonesty next to getting the description exactly right. That instinct is morally coherent. It is also, per Hobbes and per Gingrich, a losing strategy in a fight where the other side correctly identified the vocabulary as the terrain, not the map of it.

Facts were never going to be enough

There’s a reason this isn’t only a political science problem, and it’s one anyone who’s spent time near healthcare should recognize immediately: we already know, empirically, that facts and willpower are a weak lever for changing behavior, and we’ve known it long enough to build entire treatment paradigms around the alternative.

Take weight loss. In the STEP 1 trial, every participant — the semaglutide group and the placebo group alike — received the same lifestyle counseling: diet, exercise, behavioral support. The counseling-only group lost 2.4% of body weight over 68 weeks. The group that also got the drug lost 14.9% — six times more, off identical education [7]. The information didn’t change. What changed was whether the desired behavior required ongoing conscious effort or happened automatically.

Automated insulin delivery tells the same story from the other side of the same disease. Patients on multiple daily injections have to do the math themselves, every time, based on food, activity, and stress — a constant cognitive tax most people, understandably, don’t sustain perfectly. Closed-loop systems that calculate and deliver insulin automatically improve time-in-range and lower HbA1c compared to injections, trial after trial — not because patients using them suddenly understood their disease better, but because the system stopped requiring them to [8]. And in opioid use disorder, medication-assisted treatment — buprenorphine, methadone — consistently outperforms abstinence-only, willpower-based treatment on the metric that actually matters: staying alive. It reduces all-cause and overdose mortality [9]. Nobody argues this is because MAT patients have better information about the dangers of opioids than the ones white-knuckling it. They don’t need better information. They need something that lowers the amount of moment-to-moment willpower the desired outcome requires.

Behavioral science has a name for this, and a bestseller behind it: Richard Thaler and Cass Sunstein’s Nudge argues that if you want to change what people do, redesigning the choice environment beats informing people about the choice, almost every time [10]. It’s the same idea Hobbes and Gingrich already applied to language, just without the medical framing. A frame is a nudge. “Death tax” doesn’t ask a voter to learn tax policy and reason their way to opposition — it does the reasoning for them, inside the word itself, the same way a pump does math a patient used to have to do by hand. “Tax relief” doesn’t argue that taxation is a burden; it just makes not thinking so require the extra effort. The conservative messaging apparatus figured out decades ago that it was building the political equivalent of an insulin pump. Liberals have mostly kept trying to win with better patient education, in a domain where we already have the data to know that doesn’t work at scale.

That’s a strange blind spot for a coalition that otherwise prefers evidence-based medicine to folk remedies. We’d never tell a patient with obesity or opioid use disorder that better facts alone would fix it — we know that’s not how behavior change works, and we built better tools because we know it. It’s worth asking why the same coalition keeps trying to win elections with the political equivalent of a pamphlet.

Where Chomsky fits, and where he doesn’t anymore

I used to find Chomsky’s model persuasive here. In Manufacturing Consent [11] and Necessary Illusions [12], he argued that a society too free to control by force instead controls opinion by manufacturing consent — media doesn’t need to be told what to say, because ownership, advertiser-dependence, and access-to-power filters produce the boundaries on their own. Nobody has to lean on anybody. The consensus is unspoken because it doesn’t need to be spoken.

I don’t think that’s the operative model anymore, and the last two years are why. Chomsky’s version assumed an industry healthy enough to police itself quietly. What we’ve watched instead is an industry already hollowed out — nearly 3,500 U.S. newspapers have closed since 2005, more than 270,000 newspaper jobs gone with them, 213 counties now with no local news source at all [13] — get leaned on directly, by name, with instruments that don’t require reading between any lines:

Named people did this, with named leverage — a regulator, a merger approval, a legislative rescission, an owner’s phone call — against an industry too financially hollowed-out to say no the way a healthier one might have thirty years ago. There’s nothing unspoken about any of it. Chomsky’s framework explains the older, structural layer of the problem; it doesn’t have a category for a regulator threatening a license on a podcast. What we’re watching now sits closer to what political-economy researchers call media capture: direct control of nominally independent outlets through licensing, litigation, and ownership pressure, stacked on top of the structural layer Chomsky described, now that the industry no longer has the balance sheet to resist it.

It sharpens the messaging-asymmetry point rather than complicating it. Grossmann and Hopkins’s top-down-versus-coalition structure isn’t just about who coins better phrases anymore — a movement organized around centralized choke points (a handful of station groups, one regulator, one merger approval) is also just easier to squeeze through those same choke points. A coalition of interest groups doesn’t have that single point of failure, but it never had the single point of control that let Gingrich distribute one glossary to an entire party in 1990 either. The messaging-discipline problem and the capture vulnerability turn out to be the same architecture, viewed from two different angles.

The other target: anonymity itself

Institutional capture handles the concentrated channels — a handful of station groups, one regulator, one merger to hold hostage. It doesn’t touch the other place opinion actually lives now: independent blogs, pseudonymous accounts, small publications, the whole decentralized layer that exists precisely because it’s cheap and dispersed enough that no single lever controls it. If you take capture seriously as a strategy, that layer is the obvious next problem — and there’s a real, current legal track record of the tool built to solve it.

It starts, as it usually does, with porn. Since 2022, roughly half of U.S. states have passed laws requiring commercial adult-content sites to verify visitors are 18 or older, typically via government ID [19]. In June 2025 the Supreme Court upheld Texas’s version, Free Speech Coalition v. Paxton, and in doing so it didn’t just bless one law — it changed the standard. The majority applied intermediate scrutiny to a law restricting adults’ access to protected speech online [20], instead of the strict scrutiny courts had used since Reno v. ACLU in 1997, which had held that speech on the internet deserves the same high protection as print, not the reduced protection given broadcast [24]. The law firm that argued the case for the state’s opponents put out a client alert titled, plainly, “What the Supreme Court’s Age Verification Decision Could Mean for All Websites — Not Just Adult Ones” [21]. No activist wrote that headline — a corporate law firm did, telling its own clients the bar just moved for everyone.

The Electronic Frontier Foundation’s objection names exactly the mechanism at stake: age-verification systems, in their words, are “surveillance systems” that tie a user’s “most sensitive, immutable data — name, face, birthday, home address — to your online activity” [22]. Courts have long treated anonymous speech itself as protected First Amendment activity — the Supreme Court held in 1995’s McIntyre v. Ohio Elections Commission that anonymous pamphleteering is a “historically protected and socially desirable tradition of dissent,” precisely because it reduces the risk of retaliation against the author [23]. A verification regime doesn’t have to say it’s targeting anonymity to functionally dismantle it — it just has to make ID-linking a routine cost of entry to enough of the internet that it stops being unusual.

We don’t have to guess what “enough of the internet” looks like once the tool exists, because the UK is already there. Its Online Safety Act’s age-assurance enforcement began July 25, 2025, aimed at porn sites — and in that same rollout, the same month, the obligation extended to Reddit, Discord, Spotify, and X, to gate content about self-harm, suicide, and eating disorders [25]. Ofcom has already issued fines exceeding £1 million for noncompliance. The category of content the verification requirement was built to protect against turned out to be the flexible part; the ID-linking infrastructure was the durable part.

I want to be precise about what’s proven here versus what’s my extrapolation. Proven: the legal standard for restricting anonymous access to online speech has been lowered in the U.S., the technical and regulatory machinery for tying identity to online access exists and is scaling, and in the one jurisdiction further down this road than we are, that machinery was already redeployed to a second content category within a month of taking effect. Not proven, and I’m not claiming it: that anyone has used age-verification data to identify or pressure the author of a political opinion piece. That’s the inference, not the finding — but it’s an inference that only requires the next content category to be “harmful speech” or “misinformation” rather than “self-harm content,” a substitution the UK just demonstrated takes about as much effort as changing one line in a compliance filing.

If that inference is right, the two threads this piece has been pulling turn out to be one thread. Media capture goes after the concentrated channels that already have names attached to them. Anonymity erosion goes after the diffuse channel that never did — the one place where not having a Gingrich memo or a Grossmann-and-Hopkins-style central node was an advantage instead of a liability, because there was nothing to squeeze. Take that away, and the asymmetry this piece has been describing stops being an asymmetry and starts looking like a monopoly.

The actual point

None of this is an argument that liberals should lie, or that “death tax” was a good-faith description of the estate tax. It’s an argument that treating language as beneath strategy — as something you get right rather than something you win — cedes the terrain to whoever shows up willing to fight on it, and leaves you structurally unprepared for the version of the fight where the terrain itself gets bought, sued, or defunded out from under you, or where the fallback channel — anonymous, decentralized, hard to squeeze — gets legislated out of being a fallback at all. Hobbes thought controlling the vocabulary was serious enough to build a theory of sovereignty around. Gingrich thought it was serious enough to win a House majority with. Medicine figured out thirty years ago that facts alone don’t change behavior, and reached for pumps and drugs and maintenance therapy instead of better pamphlets. What’s happened to CPB, to CBS, to ABC, to two of the country’s biggest newspapers, and now to the legal architecture of anonymous speech itself, in the last twenty months, is what happens when one side stopped debating whether language is a weapon and started killing with it — and the other is still arguing about whether it’s decent to pick one up.

Sources

  1. Hobbes, Thomas. Leviathan. 1651. Project Gutenberg, www.gutenberg.org/files/3207/3207-h/3207-h.htm. Chapters 4 (“Of Speech”) and 29 (“Of Those Things That Weaken, or Tend to the Dissolution of a Commonwealth”).
  2. “Language: A Key Mechanism of Control.” GOPAC memo, 1990. Transcribed at University of Houston, www.uh.edu/~englin/rephandout.html.
  3. “USA: Memo Exposes Bush’s New Greenwash Strategy.” CorpWatch, 4 Mar. 2003, www.corpwatch.org/article/usa-memo-exposes-bushs-new-greenwash-strategy.
  4. “Frank Luntz, the GOP’s Message Master, Calls for Climate Action.” Grist, 2019, grist.org/article/the-gops-most-famous-messaging-strategist-calls-for-climate-action/.
  5. Grossmann, Matthew, and David A. Hopkins. Asymmetric Politics: Ideological Republicans and Group Interest Democrats. Oxford University Press, 2016.
  6. Lakoff, George. Don’t Think of an Elephant! Know Your Values and Frame the Debate. Chelsea Green Publishing, 2004.
  7. “STEP-1: Semaglutide 2.4 mg for Weight Management.” American College of Cardiology, 2021, www.acc.org/Latest-in-Cardiology/Clinical-Trials/2021/02/18/19/23/STEP-1.
  8. “Closed-Loop Insulin Delivery Systems: Past, Present, and Future Directions.” PMC, National Library of Medicine, pmc.ncbi.nlm.nih.gov/articles/PMC9207329/.
  9. “Medications for Opioid Use Disorder.” National Institute on Drug Abuse, nida.nih.gov/research-topics/medications-opioid-use-disorder.
  10. Thaler, Richard H., and Cass R. Sunstein. Nudge: Improving Decisions About Health, Wealth, and Happiness. Yale University Press, 2008.
  11. Herman, Edward S., and Noam Chomsky. Manufacturing Consent: The Political Economy of the Mass Media. Pantheon Books, 1988.
  12. Chomsky, Noam. Necessary Illusions: Thought Control in Democratic Societies. South End Press, 1989.
  13. “An Alarming Number of Independent Publishers and Small Chains Closed Papers Last Year, New Medill Study Finds.” Poynter, 2025, www.poynter.org/business-work/2025/medill-report-local-news-closures-independent-papers-news-deserts/.
  14. “Congress Votes to Strip More Than $1 Billion in Funding for NPR, PBS.” Axios, 18 July 2025, www.axios.com/2025/07/18/npr-pbs-funding-senate.
  15. “Paramount Agrees to Pay $16 Million to Settle Trump’s ‘60 Minutes’ Lawsuit.” CBS News, 2 July 2025, www.cbsnews.com/news/paramount-trump-60-minutes-lawsuit-settlement/.
  16. “ABC Agrees to Pay $15 Million to Trump’s Presidential Library to Settle Defamation Lawsuit.” PBS NewsHour, Dec. 2024, www.pbs.org/newshour/politics/abc-agrees-to-pay-15-million-to-trumps-presidential-library-to-settle-defamation-lawsuit.
  17. “Nexstar and Sinclair, Two Largest Station Groups, Wield Influence in ABC Decision to Pull Jimmy Kimmel in Light of His Charlie Kirk Comments.” Deadline, Sept. 2025, deadline.com/2025/09/nexstar-jimmy-kimmel-charlie-kirk-1236547382/.
  18. Murray, Conor. “Billionaire Newspaper Owners — Bezos and Soon-Shiong — Kill Kamala Harris Endorsements at Washington Post, L.A. Times.” Forbes, 25 Oct. 2024, www.forbes.com/sites/conormurray/2024/10/25/ex-editor-baron-decries-disturbing-spinelessness-of-washington-post-not-endorsing-for-president/.
  19. “Supreme Court Case on Age Limits for Porn Sites Could Affect 19 States.” Axios, 16 Jan. 2025, www.axios.com/2025/01/16/adult-website-age-verification-states.
  20. Free Speech Coalition, Inc. v. Paxton. No. 23-1122, U.S. Supreme Court, 27 June 2025. Opinion at www.supremecourt.gov/opinions/24pdf/23-1122_3e04.pdf.
  21. Huston, Michael, et al. “Free Speech Coalition v. Paxton: What the Supreme Court’s Age Verification Decision Could Mean for All Websites — Not Just Adult Ones.” Ashurst Perkins Coie, 25 July 2025. Mirrored at JD Supra, www.jdsupra.com/legalnews/free-speech-coalition-v-paxton-what-the-5028449/. (Original firm URL redirects following the Perkins Coie–Ashurst merger; mirror used for a stable link.)
  22. “The Supreme Court’s Decision on Age Verification Tramples Free Speech and Undermines Privacy.” Electronic Frontier Foundation, 27 June 2025, www.eff.org/deeplinks/2025/06/todays-supreme-court-decision-age-verification-tramples-free-speech-and-undermines.
  23. McIntyre v. Ohio Elections Commission, 514 U.S. 334 (1995). Justia U.S. Supreme Court Center, supreme.justia.com/cases/federal/us/514/334/.
  24. Reno v. American Civil Liberties Union, 521 U.S. 844 (1997). Justia U.S. Supreme Court Center, supreme.justia.com/cases/federal/us/521/844/.
  25. “UK’s Online Safety Act of 2023 in 2025.” Fordham Political Review, 2025, fordhampoliticalreview.org/uks-online-safety-act-of-2023-in-2025/.

Research and drafting assistance for this piece, including source verification, was provided by Claude (Anthropic). All claims above are cited to primary or secondary sources linked in the list above; the analysis and opinions are my own.