Why Judgment Is Ours

I wrote recently about how I divide cognitive labor with AI. Judgment on one side. Execution on the other. But what do we actually mean by judgment? What is it, where does it come from, and why does it matter who is doing it?

The cognitive scientist John Vervaeke offers a way into this. He describes human knowing as multiple faculties operating at once. There is knowing that. Facts, reasoning, conclusions. There is knowing how. The body’s intelligence, built through practice. There is knowing what it is like. The perspective shaped by every previous decision you have made in situations like this one. And there is participatory knowing. The deepest kind. You are not analyzing the situation. You are inside it, changed by it, changing it by being there.

Vervaeke calls the faculty that integrates all of this relevance realization: the ability to know what matters in a situation without having to process every piece of data to get there. It is fast, it is improbably accurate. And it is fundamentally about meaning: what is significant, what is worth attending to, what counts.

You cannot make meaning without being a subject for whom things actually matter. A subject with stakes, a body, a life. A machine can reason, weigh, and conclude. But it cannot make meaning. Nothing is at stake for it. Nothing matters to it. Its judgment, however coherent, is meaningless in the literal sense.

Judgment informed by meaning is also uncomfortable. It demands you stay present, own the call. Which is exactly why we are drawn to hand it off. AI offers a way out: the illusion of certainty, without the meaning. Instead of trusting what you know, you second-guess it. You ask the machine.

Anthropic’s research on disempowering patterns across 1.5 million AI conversations showed what this produces. Users ask “what should I do?” They accept the answer. They come back and ask again. And then the regret: “I should have listened to my intuition.” The regret is the feeling of having acted on a judgment with no meaning behind it. The decision sounded right but it wasn’t grounded in anything that mattered. The consequences fall on the human. The machine feels none of it.

The regret is not the worst part - the worst part is the habit. Each time you defer, you trust yourself a little less. Each time you trust yourself less, you defer again. It is not one bad decision but a pattern that compounds. The muscle atrophies not through a single act of outsourcing but through the accumulation of small ones.

Of course AI judges. We want it to. We want it taking things off our plate. The question is what we trust it to judge, and where we draw the line. That is not a question we can address just once. That line moves with every new capability.

I am experiencing how it is being redrawn.

I have built a Claude Skill that attempts to encode my judgment. A conciseness evaluator trained on my own standards, structures and constraints. I teach it what I can and it holds me to it. The Skill can tell me a passage fails, but it can’t tell me whether the failure matters. The encoding is always a reduction. Meaning is not.

We will increasingly be able to train our AIs on our own judgment, and AI will take on more and more. That is a tremendously useful thing. But meaning cannot be encoded. Knowing what matters, and in what context, is not something we can train into a model. Only we can be the judge of that.