Information-Scent Failure¶

Diagnose a navigation failure as the surface cues a user reads before clicking — link text, labels, snippets — failing to predict the destination's content, so the user invests a click on a wrong prediction and backtracks, hunts, or abandons.

Core Idea¶

Information-scent failure is the interaction-design failure in which the surface cues a user reads before committing to a navigation path — link text, menu labels, search-result snippets, icon glyphs, breadcrumbs, hover previews, tab titles — do not predict the destination's content well enough to guide the click toward the target. The user invests the cost of a click on the basis of the cue and arrives at content that does not match the prediction the cue created; they backtrack, spend additional clicks, or abandon the task when accumulated mismatch cost exceeds the task's value. The term derives from Pirolli and Card's information-foraging theory (Xerox PARC, 1995 onward), which models navigation as an analog of animal patch-foraging: the user allocates attention and click-cost based on the scent — the cue's predicted value — and bad scent misdirects foraging effort.

The structural failure has three components: a navigation surface presenting a set of options (menu, hyperlink list, search-results page, file browser, sitemap); a cue-text representation of each option that the user scans before committing; and a prediction gap between the cue and the destination, such that the user's click-investment is systematically misallocated. The prediction gap takes several forms — vocabulary mismatch (the label uses the designer's internal terminology where the user uses task-oriented language), specificity mismatch (a vague label over a highly specific destination, or vice versa), category mismatch (a label that announces one type of content but leads to another), or staleness (the destination was restructured but the cue was not updated to match).

The failure is distinct from discoverability failure, which is when no cue exists at all, and from empty-state failure, which is when navigation succeeds but the arrived-at view contains no content. Information-scent failure is specifically about the cue-to-destination predictive relationship: the cue exists, the destination exists, the user sees the cue, but the cue is an unreliable predictor of what is at the destination.

Structural Signature¶

Sig role-phrases:

the navigation surface — a set of options presented for choice: menu, hyperlink list, search-results page, file browser, sitemap, tab system, table of contents
the cue-text representation — the short, scannable surface cue for each option (link text, label, snippet, icon, thumbnail, breadcrumb) the user reads before committing
the scent function — the cue's power to predict its destination's content, the object that can break
the forager economics — the user as a patch-forager allocating scarce attention and click-cost against the cue's predicted value
the prediction gap — the break: the cue mispredicts the destination, factoring into vocabulary mismatch, specificity mismatch, category mismatch, or staleness
the misallocated click-investment — the user commits a click on the cue and arrives at content not matching the prediction, then backtracks, hunts, or abandons as accumulated mismatch cost exceeds the task's value
the click-and-bounce signature — the instrumented fingerprint: short dwell time followed by back-navigation, readable as a single scalar in analytics
the middle-step position — the failure sits between knowing to look and reading what was found, distinct from no-cue-exists (upstream) and unscaffolded-destination (downstream)
the gap-kind-keyed remedy — fix landing on the mismatch found: align cue vocabulary to user research (card-sort, tree-test), surface destination metadata (previews, snippets), or update stale labels

What It Is Not¶

Not discoverability failure. Discoverability failure is the absence of any cue — the user never knows a capability is there to look for. Information-scent failure is the opposite: the cue exists and the user is reading it, but it mispredicts the destination. The cue is present and attended; what is broken is its predictive power, not its existence.
Not empty-state failure. Empty-state failure occurs after navigation succeeds — the user arrives at a destination and finds it unscaffolded or empty. Scent failure is before arrival: the cue sends the click to the wrong place, or to content that does not match the prediction. One is a problem of where you land; the other, of which way the label pointed.
Not user confusion or bad taste. The diagnosis is not "users are confused" or "the menu is ugly." It is a specific, measurable structural claim — the cues do not predict the content behind them — with an instrumented fingerprint: short dwell time followed by back-navigation (click-and-bounce). The user's attention is fine; the prediction the cue created is what failed, and it is readable in analytics, not a matter of opinion.
Not a missing or broken destination. The destination exists and contains content; nothing downstream is broken. The failure is entirely in the cue-to-destination predictive relationship — the label lies about where it leads. Looking for a fault at the destination mislocates a problem that lives in the cue.
Not signaling proper. Scent cues are signals, but information-scent failure concerns only the cue-content prediction gap; it carries none of the signal-cost or equilibrium reasoning that the signaling prime adds. It is better understood as a predictive sub-property of affordance — cue-quality for predicting destinations — than as strategic signaling under cost.
Not a matter of cue quantity or visibility. The fix is not more cues, bigger labels, or more prominent placement. The cue is seen and read; the defect is in what it predicts, not whether it is noticed. A perfectly visible, prominent label that points the wrong way is a scent failure; adding visibility to a mispredicting cue does not repair the prediction.

Scope of Application¶

Information-scent failure lives across HCI, interaction design, and information architecture — every navigation surface where a user must commit a click on the basis of a cue; its reach is within that domain. The cross-domain cue-mispredicts-destination cousins (road signs, shelf browsing) are carried by signaling + attention (+ summary_substance_divergence), not as literal scent-failure habitats here, and the foraging-and-analytics apparatus does not travel to them.

Website navigation menus — top-level labels that do not predict subsection content, so users hunt across menus iteratively.
Search-result snippets — snippet text that does not predict whether a result answers the query, so users click sequentially through results.
File and folder naming — directory names that do not predict contents, so users descend the wrong subtree.
Email subject lines — subjects that do not predict whether a message needs attention, delaying triage.
Documentation information architecture — section headings that do not predict whether the answer is under them.
Library-catalogue subject headings — LCSH/Dewey classifications that mispredict a book's topic, so readers browse adjacent shelves to compensate.
Government services portals — form-finder labels that do not predict which form a user needs, so they navigate by guesswork.
Streaming thumbnails and titles — thumbnail/title scent that does not predict whether a title matches interest, so users browse extensively before committing.
Code-editor file trees and symbol lists — file names and identifiers that do not predict their content, so the developer opens the wrong file.

Clarity¶

Naming information-scent failure isolates the cue-to-destination predictive relationship as the thing that can break, separating it from the navigation failures it is constantly lumped with. Three modes get conflated under "users couldn't find it" or "the IA tested poorly," and they sit at different points in the navigation chain: discoverability failure (no cue signals the capability, so the user never knows to look), scent failure (the user knows to look and is reading cues, but the cues mispredict the destination, so the search is misdirected), and empty-state failure (navigation succeeds but the arrived-at view is unscaffolded). Holding the middle step distinct tells a designer that the user's attention is not the problem and the destination's existence is not the problem — the prediction the cue created is — and that fixes aimed at the other two stages cannot move a click path that fails because the labels lie about where they lead.

The concept's sharper work is to reframe a felt verdict — "users are confused," "the menu is bad," "the design is ugly" — as a specific, measurable structural claim: the surface cues a user reads before committing a click do not predict what is at the destination. Once the scent function is the named object, the prediction gap decomposes into diagnosable kinds — vocabulary mismatch (designer's terms vs. the user's task language), specificity mismatch, category mismatch, staleness — each with its own remedy, so "improve the navigation" becomes "align the cue vocabulary to user research, surface destination metadata, fix the stale labels." It also gives the failure an observable signature distinct from taste or opinion: a short dwell time followed by back-navigation — click-and-bounce — is the instrumented fingerprint of misallocated foraging effort, which turns the sharp question from "do users like this menu?" into "do the cues on this surface reliably predict the content behind them, and where does click-and-bounce say they do not?"

Manages Complexity¶

A UX team otherwise drowns in a heterogeneous stream of navigation complaints, each phrased as a separate verdict: "users couldn't find feature X," "the menu IA tested poorly," "the click path was longer than necessary," "search was ineffective," "the redesign feels confusing," and each spanning a different surface — top-level menus, search snippets, folder trees, documentation headings, catalogue subject headings, government form-finders, streaming thumbnails, code-editor file lists. Treated complaint by complaint, surface by surface, each invites its own redesign and its own debate about taste. Information-scent failure compresses that sprawl onto a single structural relation: between any navigation surface, the short cue-text the user scans before committing a click, and the prediction gap between cue and destination. Naming the scent function as the object means the team reasons about one thing — whether the cues on a surface reliably predict the content behind them — rather than re-diagnosing each surface and each complaint independently; every one of those scattered verdicts becomes the same predictive relationship failing on a different surface. The compression then sharpens twice. First, the prediction gap is not an undifferentiated blur but factors into a small, fixed set of diagnosable kinds — vocabulary mismatch (designer's terms vs. the user's task language), specificity mismatch, category mismatch, staleness — and each kind carries its own remedy, so "improve the navigation" resolves into a finite branch: align cue vocabulary to user research, surface destination metadata, fix the stale labels. The analyst reads which branch applies off the kind of mismatch rather than reinventing a fix per surface. Second, the failure acquires an instrumented signature that pulls it out of the register of opinion entirely: short dwell time followed by back-navigation — click-and-bounce — is the measurable fingerprint of misallocated foraging effort, a single scalar readable in analytics across every surface at once. So the qualitative outcome the team tracks collapses to two readable quantities — does the cue vocabulary match the user's task language, and where does click-and-bounce spike — and the open-ended "do users like this menu?" becomes the sharp, instrumented "do the cues on this surface reliably predict the content behind them, and where does the data say they do not?" Locating the failure precisely as the middle step in the navigation chain prunes the analysis further still: because it is cleanly separated from discoverability failure (no cue exists, upstream) and empty-state failure (navigation succeeded but the destination is unscaffolded, downstream), the team knows the user's attention is not the problem and the destination's existence is not the problem — only the prediction the cue created is — so effort is never misdirected at the wrong stage, and an entire class of navigation complaints reduces to one scent function, a fixed menu of gap-kinds, and one measurable click-and-bounce signal.

Abstract Reasoning¶

Information-scent failure licenses reasoning moves that all treat the scent function — the cue's power to predict its destination — as the object that can break, and the user as a forager spending click-cost against that prediction. The diagnostic move runs from an instrumented symptom to a determinate cause: observing a short dwell time followed by back-navigation — click-and-bounce — the analyst infers misallocated foraging effort, reasoning that the user committed a click on a cue that mispredicted what lay behind it. The reasoning is explicitly not an inference to user confusion or bad taste; from the click-and-bounce fingerprint, readable as a single scalar in analytics across every surface at once, the analyst locates the break in the cue-to-destination predictive relationship and predicts where the surface's labels are lying about where they lead. The diagnosis sharpens further by factoring the prediction gap into a fixed set of kinds — vocabulary mismatch (the designer's internal terms where the user thinks in task language), specificity mismatch (a vague cue over a precise destination or the reverse), category mismatch (a label announcing one content type but leading to another), and staleness (the destination was restructured but the cue was not) — and reasoning from the observed mismatch to which kind is present, because each kind is a different defect with a different fix.

The boundary-drawing move fixes the failure as the middle step in the navigation chain — knowing to look, then choosing where to look, then reading what was found — and that placement is load-bearing for diagnosis. Reasoning from it, the analyst separates scent failure (the cue exists and the user is reading it, but it mispredicts) from discoverability failure upstream (no cue exists, so the user never knows to look) and empty-state failure downstream (navigation succeeded but the arrived-at view is unscaffolded). The decision procedure is sharp: did the user never know to look (discoverability, place a cue), know to look and get pointed wrong (scent failure, fix the prediction), or arrive and find nothing (empty-state, scaffold the view)? Crucially the placement tells the analyst what is not the problem — the user's attention is fine and the destination's existence is fine — so it predicts that interventions aimed at those two stages cannot move a click path that fails because the labels mispredict, and effort is kept off the wrong stage.

The interventionist move reasons forward from the diagnosed gap-kind to a remedy with a predicted effect on foraging cost, each fix landing on the kind of mismatch found. A vocabulary mismatch predicts that realigning cue text to user-research language (card-sort, tree-test) will let the user's prediction match the destination; a specificity or category mismatch predicts that surfacing destination metadata in the cue — a scannable summary, a hover preview, a richer snippet — will close the gap; a staleness mismatch predicts that updating the stale labels to the restructured destination will restore the prediction. Each move carries the same directional claim: that the click-and-bounce signal should fall as the scent function is repaired, and the magnitude of the drop measures how much of the original misdirection was scent rather than something else. The move also reasons about the forager's economics over time — because users learn whether a surface's scent is reliable and adjust their click investment accordingly, the analyst predicts that a surface with reliably good scent earns cheap, confident clicks while one with bad scent earns reluctant clicks or abandonment as accumulated mismatch cost exceeds the task's value. The recurring habit the concept installs is to ask of any navigation surface not "do users like this menu?" but whether its cues reliably predict the content behind them, which kind of prediction gap is present where they do not, and where click-and-bounce says the scent has failed.

Knowledge Transfer¶

Within HCI, interaction design, and information architecture the concept transfers as mechanism, because the three-part structure — a navigation surface, the short cue-text the user scans before committing a click, and the prediction gap between cue and destination — and its foraging model apply across every surface where a user must commit a click on the basis of a cue. The diagnostic (read click-and-bounce as misallocated foraging effort, not user confusion), the gap-kind factoring (vocabulary / specificity / category / staleness mismatch, each with its own remedy), the middle-step boundary against discoverability and empty-state failures, and the interventions (align cue vocabulary via card-sort and tree-test, surface destination metadata in previews and snippets, fix stale labels, instrument click-and-bounce) all carry intact across website navigation menus, search-result snippets, file and folder naming, email subject lines as triage scent, documentation IA, library-catalogue subject headings, government form-finders, streaming thumbnails, and code-editor file trees and symbol lists. The supporting craft transfers with it: card-sort/tree-test methodology, SERP snippet and rich-result design, breadcrumbs and sitemaps as parallel scent channels, web-analytics dwell-then-bounce instrumentation, and the accessibility rule against "click here" link text (the scent-failure rule in another framing). Within the domain this is mechanism transfer, not analogy.

Beyond computing interfaces the honest report is (B) shared abstract mechanism — with the parent already split across existing primes rather than concentrated in one novel candidate. Strip the HCI vocabulary and the pattern is: a cue surface intended to predict a downstream destination or state fails to do so, so the cue's consumer invests cost on a wrong prediction. That recurs as genuine co-instances across domains — physical wayfinding (a road sign that mispredicts the next intersection), library cataloguing (an LCSH subject heading that mispredicts a book's content), abstract-versus-paper mismatch (an abstract that mispredicts the result), product labels that mispredict contents, and signage in any built environment. But the cross-domain force here does not warrant a fresh named parent, because it is substantially carried by primes already in the catalog: signaling (a cue conveying information about something not directly observed), attention (the scarce resource the forager allocates against the cue), and the sibling emergent summary_substance_divergence (a summary surface that mispredicts the substantive surface — of which a navigation cue mispredicting its destination is a close relative). So the cross-domain lesson should be carried by signaling + attention (and summary_substance_divergence where the cue is a summary), not by "information-scent failure," which on the strip-the-jargon test dissolves into exactly that combination.

What stays home-bound is the entry's HCI cargo: the scent function as a named navigational object, the click-and-bounce instrumented fingerprint, the four-way gap-kind taxonomy as a diagnostic menu, and the card-sort/tree-test/snippet-enrichment intervention family. These presuppose a designed navigation surface with clickable cues and analytics, and they do not survive extraction to a road sign or a shelf of books — adjacent substrates use their own vocabularies (wayfinding failure, subject-heading mismatch, index-term mismatch) for the same underlying gap. So invoking "information-scent failure" for a misleading road sign is (A) analogy in naming — the cue-mispredicts-destination mechanism really is shared, but the foraging-and-analytics apparatus does not come along. Two boundaries keep this exact: information-scent failure is the middle step in the navigation chain (knowing to look → choosing where to look → reading what was found), distinct from discoverability_failure upstream (no cue exists) and empty_state_failure downstream (navigation succeeded but the destination is unscaffolded); and scent is a predictive sub-property of affordance (cue-quality for predicting destinations), not the action-possibility relation itself, and is distinct from signaling proper (which adds signal-cost / equilibrium reasoning this entry does not need). When the lesson generalizes, carry signaling + attention (+ summary_substance_divergence); "information-scent failure" is their interaction-design instance, and may ultimately belong to a domain-specific cluster page alongside its navigation-chain siblings. See Structural Core vs. Domain Accent for why this profile places the entry as a domain-specific instance of those primes rather than a prime.

Examples¶

Canonical¶

The defining instance is a navigation menu whose labels use the organization's internal vocabulary instead of the user's task language — the classic case that Pirolli and Card's information-foraging theory (Xerox PARC, from 1995) models as weak scent. A shopper trying to return an item scans a top navigation reading "Customer Care," "Resources," and "My Account," and cannot tell which one leads to the return policy; the cues create no reliable prediction of where returns live. The user guesses, clicks "Customer Care," finds FAQ articles instead, backs out after a couple of seconds, tries "My Account," and either eventually stumbles onto returns or abandons the task. The labels are visible and read — nothing is hidden — but they mispredict their destinations.

Mapped back: The top menu is the navigation surface and "Customer Care"/"Resources" are the cue-text representation; their failure to point the shopper to returns is a break in the scent function, specifically a vocabulary-mismatch prediction gap (designer terms vs. the user's "returns"). The two-second visit followed by backing out is the click-and-bounce signature, the instrumented fingerprint of misallocated foraging effort rather than a matter of taste.

Applied / In Practice¶

Information architects operationalize the scent function through tree-testing (also called reverse card-sorting), a standard UX method run with tools such as Treejack. Participants are given real tasks ("where would you find the return policy?") and navigate a stripped-down, label-only version of the site's menu tree — no visual design, no search — while the tool records where they click, whether they reach the correct destination, and where they backtrack. Low task-success and heavy back-navigation on a branch localize exactly which labels mispredict their contents, and the fix is to rename them using the vocabulary surfaced by a prior card-sort of how users themselves group and name the content.

Mapped back: Tree-testing isolates the scent function directly — it strips away everything but the cue-text representation and the navigation surface, so any failure is attributable to the cue-to-destination prediction alone. The backtracking it records is the click-and-bounce signature in a controlled setting, and renaming from card-sort vocabulary is the gap-kind-keyed remedy for a vocabulary mismatch: align cue text to the user's task language rather than adding more or bigger labels.

Structural Tensions¶

T1: Middle-step failure versus its navigation-chain neighbors (the same complaint, three different stages). "Users couldn't find it" collapses three failures that sit at different points in the chain — discoverability (no cue exists, so the user never knows to look), scent (the cue exists and is read but mispredicts the destination), and empty-state (navigation succeeded but the arrived-at view is unscaffolded). The concept's diagnostic power comes from fixing scent failure as the middle step and thereby declaring what is not the problem: the user's attention is fine and the destination's existence is fine. But that placement is load-bearing in both directions — mislocate a scent failure as discoverability and you add cues that were already present; mislocate it as empty-state and you scaffold a destination the click never reaches. The tension is that the surface symptom is shared across all three stages while the correct intervention differs completely, so the whole diagnosis turns on correctly placing the break in the chain. Diagnostic: Did the user never know to look (discoverability), know to look and get pointed wrong (scent), or arrive and find nothing (empty-state)?

T2: Prediction quality versus cue visibility (making a mispredicting label louder does not repair it). The reflexive fixes for a navigation problem are to add cues, enlarge labels, or raise prominence — all aimed at whether the cue is noticed. Scent failure is defined against exactly this: the cue is seen and read; what is broken is what it predicts. A perfectly visible, prominent label that points the wrong way is a scent failure, and adding visibility to it does not close the prediction gap. The tension is that the intuitive lever (make it more salient) operates on the wrong variable, and can even worsen matters by drawing more confident clicks onto a cue that misdirects them. The concept insists the defect lives in the cue-to-destination relationship, not in the cue's conspicuousness, so effort spent on emphasis is effort spent at the wrong margin. Diagnostic: Is the cue being missed (a visibility problem) or seen-and-followed-wrong (a prediction problem) — and does the proposed fix change what the cue predicts or merely how loudly it says it?

T3: Instrumented fingerprint versus signal ambiguity (click-and-bounce pulls the failure out of opinion but does not name the stage). Short dwell followed by back-navigation is the concept's great objectifying move: it turns "do users like this menu?" into a measurable scalar readable across every surface at once. Yet the same click-and-bounce pattern is produced by more than a scent gap — an empty-state destination, a quick satisfied answer, or an idle browse can all read as short-dwell-then-back — so the metric that establishes that something is wrong under-determines which stage broke. The tension is that the fingerprint's cross-surface generality is exactly what makes it stage-blind: it localizes attention to a branch reliably but cannot by itself distinguish a mispredicting cue from an unscaffolded landing or a satisfied bounce. Reading the scalar as proof of scent failure specifically imports a conclusion the number does not carry. Diagnostic: Does the click-and-bounce here reflect a cue that mispredicted (scent), a destination that was empty (empty-state), or a need already met on arrival (satisfied bounce) — and what evidence beyond dwell-time discriminates them?

T4: Accidental prediction gap versus strategic misdirection (design defect or deliberate bad scent). The concept models scent failure as an unintended mismatch — the designer's vocabulary drifts from the user's, a label goes stale — and its remedies (realign to user research, surface metadata) assume the designer wants the cue to predict truthfully. But some cues are engineered to mispredict: clickbait thumbnails, dark-pattern labels that route users toward the provider's interest rather than the task. There the cue-to-destination gap is a feature, not a bug, and no card-sort will "fix" it because the misprediction is the goal. This is exactly the seam where the concept borders signaling proper, whose signal-cost and equilibrium reasoning the entry deliberately omits. The tension is that the instrumented diagnosis looks identical in both cases (bad scent, click-and-bounce) while the underlying situation — cooperative design defect versus adversarial misdirection — calls for opposite responses (repair versus regulate). Diagnostic: Does the party controlling the cue benefit from the user following it wrong — making the misprediction strategic — or is the gap an unintended defect they would want corrected?

T5: Repairing the scent function versus the forager's accumulated trust (behavior lags the fix). The forager economics run over time: users learn whether a surface's scent is reliable and adjust their click investment accordingly, so good scent earns cheap confident clicks and bad scent earns reluctant clicks or abandonment. This makes the failure non-memoryless in a way the click-and-bounce signal does not immediately show. A surface with a history of misprediction keeps drawing hesitant clicks even after its labels are corrected, so the metric may not fall as fast as the repair predicts; conversely, a trusted surface earns confident clicks that mask a newly introduced bad cue, because users extend it the benefit of accumulated reliability. The tension is that the intervention acts on the instantaneous scent function while the observed foraging behavior reflects a learned reputation that responds with lag. Diagnostic: Is the persistent (or absent) click-and-bounce here tracking the current cue quality, or the user population's accumulated trust in this surface built up before the fix?

T6: Autonomy versus reduction (its own HCI failure mode or the interaction-design instance of its parents). "Information-scent failure" carries real proprietary cargo — the scent function as a named navigational object, the click-and-bounce fingerprint, the four-way gap-kind taxonomy, the card-sort/tree-test/snippet-enrichment intervention family — and within HCI, interaction design, and information architecture it transfers as mechanism across menus, snippets, file trees, and thumbnails. But beyond designed navigation surfaces it does not travel as named: a misleading road sign or a mispredicting shelf heading is the same cue-mispredicts-destination structure, yet the foraging-and-analytics apparatus does not come along, and adjacent substrates use their own vocabularies. What genuinely carries is signaling (a cue conveying information about something not directly observed), attention (the scarce resource allocated against the cue), and the sibling summary_substance_divergence where the cue is a summary — with scent itself being a predictive sub-property of affordance rather than a prime. The entry itself flags that it may belong to a domain-specific cluster as an instance of those primes. The tension is between a named HCI failure mode with its own toolkit and the recognition that its cross-domain content is that parent combination. Diagnostic: Resolve toward the parents (signaling + attention, plus summary_substance_divergence for summary cues) when carrying the cue-mispredicts-destination lesson to road signs or shelves; toward the named failure when diagnosing a designed navigation surface with clickable cues and analytics in situ.

Structural–Framed Character¶

Information-scent failure sits at mixed, leaning toward the framed side of that band — closer to a framed judgment than isostasy's mixed-structural balance, because it is a named failure mode of a designed artifact rather than a mechanism nature runs observer-free. On evaluative_weight it leans framed: "failure" is in the name, and the concept exists to flag a defect that ought to be repaired, so it carries a corrective verdict a value-neutral mechanism does not — yet the underlying object it points at (a cue that mispredicts its destination) is itself a describable, measurable relation, not a bare value judgment, which keeps the evaluative charge partial rather than total. On human_practice_bound it is firmly framed: the pattern is constituted by the practice of interaction design — it presupposes a designed navigation surface, clickable cues, a user foraging under click-cost, and analytics to read the click-and-bounce fingerprint; strip that apparatus away and there is no scent function to fail. On institutional_origin it leans framed: the entry is furniture of a specific tradition — Pirolli and Card's information-foraging theory, the card-sort/tree-test toolkit, the click-and-bounce instrument, the four-way gap-kind taxonomy are all distinctions drawn inside HCI, not substrate-neutral form. On vocab_travels it is mixed-to-low: within HCI, interaction design, and information architecture the vocabulary carries its full content across menus, snippets, file trees, and thumbnails, but beyond designed interfaces the foraging-and-analytics cargo does not come along, and adjacent substrates (road signs, shelf headings) reach for their own words. On import_vs_recognize it patterns as import-by-analogy at the domain edge: the entry itself concedes that calling a misleading road sign "information-scent failure" borrows the shape while the cross-domain content is actually carried by other primes.

The portable structural skeleton is a cue surface fails to predict its downstream destination, so the cue's consumer invests cost on a wrong prediction — and, per the entry's own analysis, that skeleton is not proprietary but is what information-scent failure instantiates from a small bundle of umbrella primes: signaling (a cue conveying information about something not directly observed), attention (the scarce resource the forager allocates against the cue), and summary_substance_divergence where the cue is a summary, with scent itself a predictive sub-property of affordance. The cross-domain reach — road signs, library subject headings, product labels, abstract-versus-paper mismatch — belongs to that parent bundle, not to "information-scent failure," whose HCI-accented specifics (the named scent function, the click-and-bounce fingerprint, the gap-kind menu, the card-sort/tree-test/snippet-enrichment remedies) stay home and do not survive extraction to a road sign. Its character: a real, measurable predictive-gap mechanism — a cue that fails to forecast its destination — wrapped in a foraging-and-analytics apparatus that pins it to designed navigation surfaces, so its cross-domain cargo travels under signaling and attention while the named failure mode stays home.

Structural Core vs. Domain Accent¶

This section decides why information-scent failure is a domain-specific abstraction and not a prime — and the entry's own analysis makes the case unusually crisp, because it locates its portable content not in one novel parent but in a bundle of primes already in the catalog.

What is skeletal (could lift toward a cross-domain prime). Strip the interface vocabulary and a thin relational structure survives: a cue surface meant to predict a downstream destination fails to do so, and the cue's consumer invests cost on the wrong prediction, then pays to recover. The portable pieces are abstract — a predictor (the cue), a predicted state it stands in for but does not directly show, a consumer allocating scarce attention and cost against the cue's forecast, and a gap between forecast and reality that misdirects that investment. This skeleton is genuinely substrate-portable, which is exactly why it recurs as co-instances well beyond screens: a road sign that mispredicts the next intersection, a library subject heading that mispredicts a book's content, an abstract that mispredicts a paper's result, a product label that mispredicts its contents. But that recurrence is precisely what the entry decomposes into signaling (a cue conveying information about something not directly observed), attention (the scarce resource the forager allocates against the cue), and summary_substance_divergence (a summary surface mispredicting the substance) — with scent itself a predictive sub-property of affordance, not a structure of its own. The skeleton is the core it shares with those primes, not what makes information-scent failure distinctive.

What is domain-bound. Everything with proprietary content is HCI cargo that does not survive extraction. The scent function as a named navigational object, the click-and-bounce instrumented fingerprint (short dwell, then back-navigation, readable as a scalar in analytics), the four-way gap-kind taxonomy (vocabulary / specificity / category / staleness mismatch) as a diagnostic menu, and the card-sort / tree-test / snippet-enrichment intervention family all presuppose a designed navigation surface with clickable cues and an analytics layer to instrument. The decisive test: remove the designed, clickable, instrumented surface — carry the concept to a misleading road sign or a mispredicting shelf heading — and none of that apparatus comes along; the cue-mispredicts-destination mechanism is genuinely shared, but the adjacent substrate reaches for its own vocabulary (wayfinding failure, subject-heading mismatch, index-term mismatch) because the foraging-and-analytics machinery is gone. What is left after extraction is the bare predictive gap, which is the parents' content, not this entry's.

Why this does not clear the prime bar. A prime's vocabulary travels and its transfer is recognition of the same mechanism, not analogy. Information-scent failure's transfer is bimodal. Within HCI, interaction design, and information architecture it travels as mechanism — the diagnostic, the gap-kind factoring, the middle-step boundary against discoverability_failure (upstream, no cue exists) and empty_state_failure (downstream, the destination is unscaffolded), and the whole intervention toolkit carry intact across menus, snippets, file trees, email subjects, catalogue headings, and streaming thumbnails, because every one supplies a designed surface with clickable cues and analytics. Beyond designed interfaces it travels only by analogy: calling a misleading road sign "information-scent failure" borrows the shape while the operative apparatus stays home. And when the bare structural lesson is needed cross-domain, it is already carried, in more general form, by the parents the entry instantiates — signaling and attention (and summary_substance_divergence where the cue is a summary), with scent as a sub-property of affordance. The cross-domain reach belongs to that parent bundle; "information-scent failure" is their interaction-design instance, carrying navigation-and-analytics baggage that should stay home — which is why the entry itself flags that it may ultimately belong on a domain-specific cluster page beside its navigation-chain siblings rather than standing as a prime.

Relationships to Other Abstractions¶

Current abstraction Information-Scent Failure Domain-specific

Parents (2) — more general patterns this builds on

Information-Scent Failure presupposes Discoverability Domain-specific

Information-Scent Failure presupposes the graded Discoverability property whose encountered cue sends exploration away from the intended destination.
Information-Scent Failure is a decomposition of Information Scent Prime

Information-Scent Failure is the HCI navigation form of a broken Information Scent relation between a local cue and a distal destination.

Hierarchy paths (8) — routes to 7 parentless roots

Information-Scent Failure → Discoverability → Coverage / Reachability → Completeness

Show alternative paths (7)

Not to Be Confused With¶

Discoverability failure. The upstream neighbor in the navigation chain: no cue exists at all, so the user never knows a capability is there to look for. Scent failure is the opposite — the cue does exist and the user is reading it, but it mispredicts the destination. Mislocating a scent failure here leads a team to add cues that were already present. Tell: did the user never know to look (discoverability) or know to look and get pointed the wrong way (scent)?
Empty-state failure. The downstream neighbor: navigation succeeds, the user arrives, and the destination is unscaffolded or empty. Scent failure is before arrival — the cue sends the click to content that does not match the prediction. One is a problem of what you find on landing; the other, of which way the label pointed. Tell: did the click reach a real, populated destination that the label simply mispredicted (scent), or reach the intended place only to find nothing there (empty-state)?
Affordance. The broader action-possibility relation — what a surface signals a user can do with it. Scent is a predictive sub-property of affordance: cue-quality for predicting destinations specifically, not the whole action-possibility relation. Part-vs-whole: scent failure is one way an affordance can be defective (the cue is legible as clickable but lies about where it leads). Tell: is the defect that the user cannot tell an action is available (broader affordance), or that they can act but the cue mispredicts the result (scent)?
Summary–substance divergence. The sibling emergent where a summary surface fails to predict the substantive surface it stands for (an abstract that mispredicts its paper, a headline that mispredicts its article). A navigation cue mispredicting its destination is a close relative — and where the scent cue is a summary, the two overlap — but scent failure is keyed to the click-and-forage economics of a designed navigation surface, not to summary/substance mismatch in general. Tell: is the mispredicting cue a condensation of the target's content (summary–substance divergence), or a navigational pointer the user clicks to forage (scent failure)?
Signaling proper. The parent prime the entry instantiates: a cue conveying information about something not directly observed. Scent failure borrows the cue-predicts-hidden-state structure but carries none of signaling's signal-cost or equilibrium reasoning — it treats the cue as a cooperative predictor that happens to break, not a strategic emission under cost. Tell: does the analysis turn on the cost of sending the cue and the sender's incentives (signaling), or purely on whether the cue predicts its destination (scent)? (The umbrella is treated more fully in a later section.)
Clickbait / dark-pattern misdirection. The contrast case — a cue engineered to mispredict on purpose, routing the user toward the provider's interest rather than the task. It produces the identical instrumented fingerprint (bad scent, click-and-bounce), but the gap is a feature, not a bug, so no card-sort will "fix" it. This is exactly the seam where scent failure borders signaling proper. Tell: does the party controlling the cue benefit from the user following it wrong (strategic misdirection), or is the misprediction an unintended defect they would want corrected (scent failure)?

Neighborhood in Abstraction Space¶

Information-Scent Failure sits in a crowded region of the domain-specific corpus (26^th percentile for distinctiveness): several abstractions share nearly its structure, so a description that fits it tends to fit its neighbors too.

Family — Surface Form & Underlying Structure (23 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-07-12