The Tyranny Dictionary Test

(and What the Algorithm Revealed)

Note to readers: This post was drafted with the assistance of AI, based on my original ideas, a Facebook prompt, the replies people shared, and my editorial direction. I reviewed and shaped it before publishing. I’ve also replaced the names of those who responded with initials, to protect their identity.

For years, I dismissed the idea that the federal government could be tyrannical in its nature.

My anti-establishment friends would tell stories about IRS enforcement or EPA rules they didn’t want to follow. In their telling, “the system” was always the villain and they were always the victim. I’d nod at the flaws, then default to my usual conclusion: yes, the system is imperfect, but it works for most people—so play along. I confess I believed they were increasing their own suffering by refusing to comply.

Now I’m the one using the word tyranny. And I don’t say that lightly.

Nonviolent civil resistance, non-compliance, and a general strike are not casual tools. They’re among the most powerful tools citizens have—and they only make sense when the situation truly warrants them. So before I write a longer piece making claims about what we’re living through, I decided to do something simpler.

I posted a challenge on Facebook:

Open a browser. Type: “What is a tyrannical government?”
What shows up first for you?
Drop it in the comments—copy/paste the first definition or describe what you saw.
Then let’s talk: tyranny… or not?

I added one sentence that mattered more than I expected:

Because algorithms decide what we see first, the definition you meet may say as much about you as it does about the word.

I thought I was running a dictionary experiment. What I actually ran was an algorithm experiment.

What people found wasn’t “the dictionary.” It was a machine-made template.

A striking number of replies shared nearly identical language—so similar it felt like people were looking at the same screen, even though they weren’t.

JR posted an AI summary that described tyranny as “absolute, oppressive, and unjust power,” marked by “abuse of power,” “lack of accountability,” and “suppression of dissent.” RS and MB posted definitions that echoed the same structure: a ruler or group exercising “absolute, unrestricted power,” acting for “their own benefit,” enforcing “cruel or arbitrary rule,” and violating legal limitations.

JL’s result—labeled “Search Assist”—hit the same notes: concentrated power, oppressive practices, disregard for rights, maintaining control through fear and suppression of dissent.

If this were truly a “dictionary” moment, we would have seen more variation: different word choices, different emphases, different historical context. Instead, many people encountered what I’d call the AI Overview Definition—a smooth paragraph that feels authoritative, and almost always includes the same handful of traits.

This is not necessarily wrong. The traits listed are common markers of tyranny across history. But when an algorithm hands us a tidy paragraph, it does something sneaky: it makes us think we did research when we mostly did retrieval.

A machine served an answer shaped like certainty.

Some people were pushed into a different frame entirely.

VL ran the experiment in multiple browsers and search engines. His first result—Google/Chrome—was from a test-prep site (Fiveable) and it included a new element that most others didn’t see: it linked the concept of tyranny directly to the Second Amendment as a protective measure.

That matters.

Because it means two people can ask the exact same question and be handed two different implied next steps. One person gets “tyranny = oppression and censorship.” Another gets “tyranny = reason for armed self-protection.” Same word. Different doorway. Different downstream politics.

That’s not a minor glitch. That’s the architecture of modern persuasion.

Then came the leap: definition → diagnosis.

Several commenters didn’t just share a definition. They immediately declared the conclusion.

SB went past the AI assistant and found a definition on a legal-sounding website, then quoted its explanation of how tyranny operates—controlling information through censorship and suppression of independent media—and concluded: “We are absolutely there.”

KB summed up what she saw as concentrated power and restrictions on speech, then added: “Like we have now, alas.”

Those are not ridiculous conclusions. They might even be correct. But they’re not the same thing as a definition.

A definition tells you what a word means.
A diagnosis tells you whether your society meets the threshold.

And “threshold” is the core question we keep skipping.

Because it’s emotionally satisfying to say, “That’s us.”
It’s also emotionally satisfying to say, “No it isn’t—people are overreacting.”
What’s harder is to ask: What evidence would make the label true? What evidence would make it false? What evidence would make it “not yet”?

One commenter refused the leap—and that refusal is part of the data.

SRC pushed back, and I’m grateful she did.

She said she saw definitions “as in actual definitions,” and that while the federal government has too much power, we aren’t experiencing tyranny. Then she pointed to something many of us don’t want to face: the internet makes it easy to spread chaotic, conspiratorial thinking. Pair that with weak education and little real understanding of historical tyrants, plus the human mind’s tendency to find patterns, and people can start calling “tyranny” whenever outcomes don’t go their way. She even used the phrase “ideological subversion tactics.”

I don’t agree with every framing in that comment—but it’s a serious argument, and it matters because it names the hazard of our time:

We can be harmed by government overreach.
And we can be harmed by information systems that turn every conflict into apocalyptic certainty.

Sometimes those harms overlap. Sometimes they’re mistaken for each other. And the algorithm profits either way.

The weird replies were also part of the experiment.

One commenter posted nonsense—dinosaur T-Rex and riddle-like fragments. That may be trolling. It may be a mind in distress. It may be someone trying to derail the conversation.

But it’s still a real feature of the public square now: serious civic questions don’t get to remain serious for long. Noise rushes in. Attention fragments. People who might have offered thoughtful nuance quietly back away.

That, too, changes what a society can talk about—and therefore what it can solve.

So what did we learn?

We learned that most people aren’t being directed to a dictionary. They’re being directed to a summary. A synthesized paragraph. A frictionless answer.

We learned that the first result can carry hidden ideological freight—like tying tyranny to weapons—or can carry emotional freight—like immediately invoking Hitler, Stalin, or Nazi Germany, as one AI overview did. (Sometimes that’s helpful context. Sometimes it functions like a rhetorical shortcut: “If it isn’t that, then it isn’t tyranny.”)

We learned that many of us are treating a definition like a verdict.

And we learned that different people, with different histories and different media diets, can read the same definition and land in opposite places. One person says, “We are absolutely there.” Another says, “We are not—and the internet is training you to think you are.”

That’s the real fight inside the word tyranny: not just what it means, but what it takes to prove.

Here’s the question I’m sitting with now.

If tyranny is “oppressive power without meaningful restraint,” then the civic task isn’t just to shout the label. The civic task is to identify the restraint mechanisms:

• Do laws apply equally—or selectively?

• Do institutions still function independently—or only when convenient?

• Are people free to dissent without fear of retaliation—or quietly self-censoring?

• Can citizens still correct course through accountable processes—or are those processes becoming performative?

Those aren’t partisan questions. They’re diagnostic questions. And they require something the algorithm doesn’t promote: patience, comparison, and the willingness to be wrong.

So before you argue with your neighbor about whether we are “in tyranny,” try this:

Ask the question again—but don’t stop at the first result. Screenshot the first thing you’re shown. Then look up an actual dictionary entry. Then read one historical source. Then come back and tell the truth about what you found. When a machine hands you certainty in one paragraph, your job is to earn your own conclusion the hard way.

Not the truth you were already hoping to find.