Thursday morning, February 26th. A researcher is asked, at a conference dinner, what she thinks about a contested empirical claim in her field—one she's studied closely and holds a clear private view on. She gives the careful answer: "The literature is genuinely mixed, there are thoughtful people on both sides, and I think we need more research." She receives appreciative nods for her nuance. She's right that the literature is mixed. She is not right that she doesn't know which way it goes. The careful answer and the honest answer were different things, and she chose the careful one.

Remaining conspicuously undecided on contested questions is treated as intellectual humility. It's usually the opposite—a social strategy that preserves optionality at the cost of honest engagement. There is real uncertainty in the world, and genuine epistemic humility is a genuine virtue. But most "I can see merit on both sides" is not that. It's the maintenance of strategic ambiguity by people who do, in fact, have views—and are choosing not to share them. The cost of this choice is paid by everyone around them who is actually trying to figure out what's true.

The Performance of Uncertainty

There's a difference between not knowing and performing not-knowing.

Real uncertainty is cognitively uncomfortable. It means holding a question genuinely open, sitting with the discomfort of ambiguity, being honestly unsure which way the evidence points. This is sometimes the right epistemic state—when evidence is genuinely mixed, when the question is genuinely hard, when you've done the work and remain uncertain.

But there's a second variety of uncertainty that's quite comfortable: the strategic kind. The person who is strategically uncertain can remain friends with everyone. They're never caught being wrong, because they never committed. They signal sophistication by acknowledging complexity. And they offend nobody, because they haven't taken a side.

This strategic uncertainty has real social value—it's how intellectuals, mediators, and anyone who needs to function across factional lines navigates contested terrain. But it's a social strategy, not an epistemic virtue. When it gets relabeled as intellectual humility, it does something dishonest: it presents a social calculation as a belief about the evidence.

The tell is what happens when the audience changes. If you'd give a different answer at a different dinner table, you're not doing epistemics—you're managing optics.

What You Lose When You Don't Commit

The feedback loop of good reasoning requires commitment to actual positions.

Philip Tetlock's superforecasters—the best human predictors in his research—share a consistent trait: they make specific, falsifiable predictions and track their accuracy obsessively. They know their error rates in different domains. They can tell you when they're well-calibrated and when they drift. They get better over time in measurable ways.

This only works because they commit. A prediction expressed as "it's complicated" teaches you nothing when it turns out to be wrong. You can't update on a non-prediction. The person who never takes a stand never discovers where their reasoning fails—because failures require the prior existence of a specific claim to fail.

Epistemic cowardice is thus self-reinforcing: by never committing, you never get feedback; without feedback, you can't calibrate; without calibration, you can't distinguish actual uncertainty from social uncertainty. The permanently undecided person doesn't know their own error rate. Their caution looks like wisdom but functions like noise.

Consider what this means at scale. When intelligent, well-informed people systematically decline to share their actual assessments, the epistemic environment is impoverished for everyone. The people who will commit—and who therefore do get corrected and do improve—are often not the most informed. The most informed people have retreated into nuance theater. The result is an intellectual culture where the confident are often wrong and the sophisticated are often invisible.

Intelligence Makes This Worse

Here's the counterintuitive part: the people most prone to strategic uncertainty are often the most analytically capable.

A sharper mind can construct a stronger case for any position—including the position of not having one. "I can see the merits of approach A and approach B" is more convincing when you can actually generate those merits on demand. High intelligence provides better tools for rationalizing whatever stance is socially useful, including the stance of deliberate non-commitment.

This is the galaxy-brained failure mode: constructing an elaborate, internally consistent case for strategic vagueness that is more sophisticated than the straightforward reasoning that would have produced an actual answer. The more capable the reasoner, the more convincing the construction. Smart people are often the last to notice when sophistication has replaced thinking.

The people least prone to this failure are often those with less analytical flexibility—they can't construct convincing arguments for every position simultaneously, so they end up committed to one. This is not intellectual virtue on their part; it's a capacity constraint. But the practical output (a committed, falsifiable position that can generate feedback) is sometimes better than the output of the strategically sophisticated non-committer.

The Social Function of Vagueness

It's worth being honest about why epistemic cowardice exists: it works.

Being visibly undecided maintains relationships across factional lines in ways that taking a stand destroys. The academic who won't say whether X is real preserves friendships with people on both sides of the debate. The consultant who "acknowledges the merits of multiple frameworks" stays employable across client types. The public intellectual who describes every controversy as "more nuanced than it appears" never has to apologize for being wrong.

There are contexts where diplomatic ambiguity is the right call. You don't share every view in every setting. Picking battles is rational. And genuine uncertainty deserves expression as genuine uncertainty.

But when the pattern is systematic—when someone is reliably and visibly undecided on exactly the questions where having an opinion would cost them something—the social calculus has colonized the epistemic one. The performance of open-mindedness has replaced the practice of it.

What Genuine Uncertainty Looks Like

The alternative to epistemic cowardice is not fake confidence. Genuine uncertainty deserves genuine expression. The markers that distinguish real epistemic humility from the performance of it are concrete.

Real uncertainty looks like: "I've studied this carefully and I genuinely don't know—here's what I'm uncertain about, here's what evidence would resolve it, here's what I provisionally believe." You can specify the location of your uncertainty. You can state what evidence would change your mind. You can give a tentative answer with honest hedging.

"It's complicated" fails all these tests. It states nothing, predicts nothing, and cannot be updated. It's not the expression of uncertainty—it's the expression of a desire not to express anything.

The practical test: if you hold a private opinion different from your public one, you're not being epistemically humble. You're being strategically ambiguous, which is sometimes socially necessary but is not an intellectual virtue. Knowing which you're doing is the beginning of doing better.

The Takeaways

Have actual opinions. On contested questions in your domain, stake out tentative, revisable positions and be willing to defend them. The hedging is legitimate—positions should be revisable—but the position itself must exist. Permanent openness to updating is not a virtue if you never had a position to update from.

Track your predictions. The feedback loop requires commitment. Write down what you expect to happen and why. Check back when you know the answer. You cannot improve reasoning you never actually exercised.

Apply the audience test. Ask whether your level of certainty changes depending on who you're talking to. If so, you're not calibrating to evidence—you're calibrating to the room. The thing to adjust is your willingness to speak honestly, not the content of what you believe.

Recognize what strategic vagueness costs others. Every time a well-informed person gives a carefully balanced non-answer to a question where they have a real view, they remove one data point from everyone trying to figure out what's true. The aggregate effect of an intellectual culture that rewards vagueness is that genuine signal about what informed people actually believe gets swamped in diplomatic noise.

Reserve "it's complicated" for when it's actually complicated. That qualifier exists for good reasons—some questions genuinely resist clean answers. But if "it's complicated" is your default response to every contested question, it's doing work that "I don't want to say" should be doing instead.

The person who never commits is never provably wrong. They are also never genuinely useful to anyone trying to think clearly about hard questions. Being specific, being falsifiable, being willing to be corrected—this is how you contribute to collective reasoning. The alternative is a form of freeloading: benefiting from the reputation of expertise while offering only the performance of thought.

Today's Sketch

February 26, 2026