“How Is That Different?” Is a Lazy Question

“How is that different?”

It sounds analytical. Comparative. Thoughtful.

It is usually none of those things.

Most people use that prompt as a reflex. The model explains one idea. They ask how it differs from another. The model lists distinctions. Different features. Different use cases. Different philosophies.

It feels like clarity.

It is often shallow contrast.

The weakness of “How is that different?” is that it assumes difference is the most important dimension. It reduces comparison to separation. It treats ideas like consumer products on a shelf: this one has feature A, that one has feature B.

But serious distinctions are not about surface variation. They are about consequences.

If you ask how two economic policies differ, the model might tell you one emphasizes markets and the other regulation. That is descriptive. It is not decisive.

The better question is: what does each one make more likely? Who gains power under each? What risks increase? What behaviors are rewarded?

Difference without impact is trivia.

There is another problem. The prompt does not define the axis of comparison.

Different in what way? Mechanism? Outcome? Cost? Risk? Time horizon? Moral framework? Operational complexity?

Without specifying the dimension, the model defaults to generic contrasts. It produces safe distinctions that cannot offend and rarely illuminate.

You get a list.

Lists are tidy. They are not sharp.

High-level operators do not ask “How is that different?” in isolation. They anchor the comparison to stakes.

“Under resource constraints, how would these approaches diverge in outcome?”
“In a high-risk environment, which model becomes fragile?”
“What assumption does one require that the other does not?”

Now the difference matters.

The original prompt is weak because it avoids commitment. It does not require you to declare what you care about. It leaves the comparison floating.

And when comparison floats, so does judgment.

There is also a subtle psychological comfort in that phrasing. “How is that different?” allows you to remain neutral. You are not evaluating. You are not choosing. You are just observing distinctions.

But most decisions do not hinge on difference alone. They hinge on superiority under constraint.

If two tools differ but one exposes you to catastrophic downside, that difference is not symmetrical. It is decisive.

The shallow comparison format flattens that asymmetry.

Language models are capable of far deeper comparative reasoning. They can map trade-offs, stress-test assumptions, and project divergent outcomes. But they will only do that if you force them to.

Otherwise, they will comply with the minimal request.

“How is that different?” invites minimal compliance.

It produces a catalog of contrasts.

It does not produce judgment.

If you want clarity that changes action, stop asking how things differ in general.

Ask how those differences alter risk, power, cost, and consequence.

Because difference without impact is decoration.

And decoration does not guide decisions.