Curiosity as an information gap

Why does knowing a little about a subject make you far hungrier to learn than knowing nothing?

There are whole fields you have never felt the slightest pull toward. Not because you decided they were dull — you have simply never thought about them. Then a friend explains one thing about one of them over dinner, and by the next morning you have read three articles.

Nothing about the field changed. Your knowledge went up, not down. If curiosity were hunger for what you lack, shouldn't it have been strongest when you knew nothing at all? Something about that model must be wrong.

Let us take the hunger analogy seriously for a moment and see where it fails. Hunger is driven by a deficit against a set point: the further below the level, the stronger the drive. Applied to knowledge, that predicts curiosity is highest at total ignorance and falls monotonically as you learn. Does that match anything you have experienced? It does not match the dinner-party case at all.

So try a different quantity. What if curiosity is not driven by the level of what you lack, but by a perceived difference — a gradient between what you know and a reference point of what you feel you could or should know? Gradients behave differently from levels in one decisive way: a gradient needs two ends. Deficit is measured against a fixed set point that exists whether or not you notice it. A gradient exists only where both ends are represented.

Now the puzzle resolves. Before your friend spoke, you had no representation of the far end. There was no felt gap, because a gap between something and nothing you have ever conceived of is not a quantity the mind computes. The first fact did not close a gap; it created one, by supplying the reference point against which your ignorance became visible and measurable. This is essentially George Loewenstein's 1994 reinterpretation of curiosity: an aversive feeling arising from an information gap between what one knows and what one wants to know.

Push the logic to the other end. If curiosity tracks the gap, what happens as you learn more? The gap narrows, and the pull should fade — which is why a subject you have genuinely mastered stops nagging at you. So the relationship between knowledge and curiosity is not a slope at all but an arch: near zero when you know nothing, near zero when you know everything, and greatest somewhere in the middle. The shape is what the deficit model could not produce, and it is the signature prediction of the gap model.

Has the arch been observed? Reasonably well. Kang and colleagues, in 2009, asked people trivia questions, had them rate their curiosity and their confidence in the answer, and found curiosity peaked at intermediate confidence — when a person felt they almost knew. High curiosity also predicted better recall of the answer later and a willingness to spend limited tokens, or wait, to find out. So the gap does not merely correlate with interest; it behaves like a motivational pressure that people will pay to relieve. I should be careful here: the arch is the reliable shape, but its exact peak varies with the person and the material, and no one has a formula for it.

One more consequence, and it is the practical one. If the pull is a gradient rather than a level, then curiosity can be created by moving either end. Add a little knowledge and the gap opens from below — the dinner-party effect. Or make the far end vivid without granting it: pose the question first, show the surprising result and withhold the mechanism, reveal that an answer exists and is knowable. A cliffhanger is not adding information; it is installing a reference point.

Which also explains a familiar teaching failure. A lecture that opens by defining terms in order supplies no far end at all, so there is nothing for attention to flow toward. The same content, opened with a question the audience cannot yet answer, has a gradient across it from the first minute.

The first refinement is that "knowing a little" is not the same as "having a little information dumped on you." What opens the gap is a representation of a specific answerable question. A vague sense that physics is complicated does not create a gradient. Learning that a spinning object resists being tipped, and not being told why, absolutely does.

The second is that the gap model does not exhaust curiosity. Berlyne's older work distinguished specific curiosity — the itch for one particular missing fact — from diversive curiosity, the general appetite for novelty and stimulation, and the gap account handles the first far better than the second. Wanting to know the ending of a story is a gap; idly browsing is closer to seeking arousal.

The third is that the gap must be reachable. A gradient so steep that the far end feels unattainable produces avoidance rather than pull, which is the same insight teachers reach for when they aim material just beyond current competence. Interest lives in a narrow band: far enough away to be a gap, near enough to feel closable.

• Whether the same arch describes what people find interesting to read versus what they choose to study for years.
• The relationship between the information gap and the tip-of-the-tongue state, where the gap is at its narrowest and most maddening.
• How question-first teaching designs compare with definition-first ones on retention, not just engagement.

Every term the collection defines is gathered in the glossary.