An Integrative Appraisal Model of Epistemic Curiosity

A 2025 review from David Sander's Geneva affective-science group that unifies three rival theories of curiosity — information-gap, emotion, and reward-anticipation — into a single appraisal-based model. Its core claim: knowledge gaps trigger curiosity only when appraised as relevant and cope-able; when relevance is high but coping potential low, the same gap produces anxiety and information-avoidance instead. The model maps each step of curiosity onto a brain circuit and ends by reframing curiosity as simultaneously an emotion and a reward-anticipation process.

Key Quotes

"If the detection of a knowledge gap was enough to elicit a desire to fill the gap, encyclopedias would be addictive."

The paper's sharpest line and its real contribution. Prior theories (information-gap, reward) explain how curiosity feels or is reinforced once a gap is noticed, but not which gaps get noticed. The authors hang the whole model on this hole: the answer is appraisal — a gap becomes curiosity only after passing a set of automatic subjective evaluations (novelty, relevance, valence, coping potential). This is the same move appraisal theory made for emotion in general, imported to curiosity specifically.

"Epistemic curiosity i) is an emotion which is triggered by appraised high relevance, valence, novelty and coping potential of knowledge-gaps; ii) confers a reward value to its epistemic object; iii) triggers a reward anticipation process; iv) and motivates a knowledge-seeking behavior (i.e., 'wanting knowledge')."

The abstract's four-part summary. Note the deliberate two-sidedness baked into the syntax: curiosity is an emotion (i) and a reward-anticipation process (iii). The authors later concede this may be "a question primarily raised by differences between research traditions than an empirical question as such." The unification is achieved partly by defining the battle out of existence.

"One might want to find out about their spouse's infidelity, if they feel like that knowledge will allow them to end the relationship or repair the relationship. If divorce is impossible because of economic difficulties, one might fear finding out about an affair and rather avoid such knowledge."

The model's most memorable illustration, and the hinge point of the whole theory: the same knowledge flips from sought to avoided depending on one appraisal dimension — coping potential. This is what distinguishes IAMEC from pure information-as-reward accounts, which would naively predict the information is always rewarding. It also explains morbid curiosity, gossip, and true-crime fascination: unpleasant knowledge is still sought when the seeker believes they can handle it.

"Making a prediction about the not-yet-acquired knowledge adds more self-relevance to the appraisal process (i.e., 'was I right or wrong in my guess?'), thus enhancing curiosity."

The paper explains a robust empirical finding — that prompting a guess before revealing an answer boosts curiosity — through appraisal: prediction injects self-relevance. It's an example of the model earning its keep on a result reward-based accounts can't easily explain, since the reward value of the information is unchanged by whether you predicted it.

Key Themes

#concept IAMEC — The integrative appraisal model of epistemic curiosity. A box-and-arrow synthesis: knowledge gap → appraisal (Scherer's Component Process Model) → reward anticipation ("wanting") → knowledge-seeking → knowledge consumption ("liking") → affective prediction error → updated expectations. Explicitly framed as falsifiable, with two named disconfirming conditions.

#concept Appraisal as the missing gate — The authors borrow Scherer's four appraisal categories (relevance, motivational implication, coping potential, normative implications) to specify why some gaps are curious and others ignored. Coping potential is the load-bearing dimension: it decides whether a gap becomes curiosity or anxiety.

#concept Wanting vs. liking — Berridge's reward distinction imported to knowledge: curiosity is "wanting knowledge" (dopaminergic, incentive salience, SN/VTA and OFC), while knowledge consumption is "liking" (hedonic hotspots in nucleus accumbens shell, ventral pallidum). The two are separate circuits and can diverge — which is precisely the mechanism Dopamine Fracking describes being industrialized.

#concept Affective prediction error — After knowledge is consumed, felt pleasure is compared to expected pleasure (an "information prediction error"); the gap updates future expectations and, the authors argue, adds surprise-driven memory enhancement. This is the loop that turns single curiosity episodes into sustained interests.

#person David Sander — Swiss affective neuroscientist (University of Geneva), co-author here and architect of the appraisal-and-reward convergence the paper leans on (Sander & Nummenmaa 2021). The group's signature move is bridging emotion science and reward neuroscience; IAMEC is that bridge applied to curiosity.

#pattern Unification by redefinition — The paper resolves the "is curiosity an emotion or a reward process?" debate by asserting it's both, then arguing the question was a research-tradition artifact. It's an honest move, but the honesty doubles as an admission that the "integration" is partly a diplomatic settlement rather than a new empirical result.

Critical Analysis

The encyclopedia line is the paper's real gift. Most integrative reviews succeed by finding an umbrella; this one succeeds by finding a genuine explanatory hole — no prior theory says why only some gaps elicit curiosity — and filling it with a specific mechanism (appraisal). That's what makes IAMEC testable rather than merely tidy. The two explicit falsification conditions (find curiosity from low-relevance/low-coping gaps; find avoidance after a positive affective prediction error) are rarer in this genre than they should be.

But "both an emotion and a reward process" is a fudge, and the authors half-admit it. When they write that the emotion-vs-reward question "may well be a question primarily raised by differences between research traditions than an empirical question as such," they're conceding that the unification is achieved by redefining the dispute as non-empirical. That's a legitimate integrative move — Sander has argued elsewhere that emotion and reward are conceptually overlapping — but a skeptic will note that a model claiming to settle a debate by asserting both horns is the debate, not its resolution. The harder unification, integrating computational accounts (active inference, learning progress) that actually make numeric predictions, is deferred to a future-direction paragraph.

The coping-potential hinge is the most useful and most under-specified part. The claim that curiosity vs. anxiety turns on a single appraisal dimension is elegant and clinically consequential (it reframes information-avoidance as failed coping appraisal, not irrationality). But "coping potential" is doing enormous work — it's a folk-psychological category smuggled into a supposedly mechanistic model, and the paper doesn't specify how it's computed or measured beyond citing the Component Process Model's "control, power, adjustment" checks. The spouse-infidelity example is vivid because it's a story, not a mechanism.

For this wiki, the wanting/liking split is the bridge. The IAMEC is the clean-slate neuroscience of the same circuit Dopamine Fracking names from the extraction side: reward anticipation that can be hijacked independently of the liking it was built to serve. Where German S. diagnoses the industrial abuse of wanting-without-liking, this paper supplies the underlying two-circuit account and even the update rule (affective prediction error) that an engagement economy would need to keep expectations ratcheting upward. It's also the mechanism underneath The Usefulness of Useless Knowledge: Flexner argued institutionally that non-instrumental knowledge pays off; IAMEC shows mechanistically why "knowledge for knowledge's sake" is a stable, self-sustaining drive rather than a luxury — the affective prediction error that makes curiosity compound.

The model's scope is quietly narrow. "Epistemic" is doing heavy lifting — the authors confine themselves to knowledge-for-knowledge, excluding perceptual, social, and instrumental curiosity, and admit (in the future directions) they don't know how the model extends to those. That's a defensible carve-out, but it means the headline "integrative model of curiosity" is really an integrative model of one curiosity. The grand synthesis is more of a well-fenced district than a unified country.

Sources: raw/s42761-025-00328-7, An Integrative Appraisal Model of Epistemic Curiosity Last updated: 2026-08-25