A long-unfilled explanation gap pushes users to fill it themselves with a wrong causal story

What it is

An explanation gap doesn't stay empty forever. If a system repeatedly does something puzzling and never gets an official attribution for it, a user won't stay indefinitely at "I don't know why" — they will invent a story of their own to fill the blank. This self-generated, often inaccurate causal explanation is a folk theory. This entry is about exactly that transition — the gap going from "temporarily empty" to "filled with the user's own guess" — and about what changes once that transition has happened: the situation is no longer simply a missing explanation.

Why it happens

A user's tolerance isn't indefinite. In the short term, an unfilled gap is just a temporary lack of attribution. But if the same kind of anomaly keeps recurring with no attributable cue ever coming from the system, the user will start their own attribution process to fill that spot — even if the resulting story has never been verified, as long as it gives them something usable for predicting and reacting to the next occurrence, it sticks around. Once a folk theory forms, it goes through round after round of the user's own confirmatory use: every time the user adjusts their behavior according to this self-made theory and it isn't disproven, the theory gets reconfirmed once more, gradually hardening from a tentative guess into a belief that's been used repeatedly and is now treated as reliable.

By this point, the explanation gap itself is no longer the core of the problem — the user is no longer "missing an explanation"; they have already filled that spot with a theory they came up with themselves, one that has already been used and confirmed by their own repeated experience. This is exactly the starting point where a wrong mental model goes from nonexistent to a loose guess to an entrenched belief: looking backward, the window before this point is where a timely, well-anchored explanation can still work; looking forward, once a folk theory has hardened through repeated confirmatory use, simply supplying, after the fact, the explanation that should have been given at the start is no longer enough to shake a theory that has already been verified many times over and that the user is now genuinely confident in — because what needs to be overcome is no longer a blank, but a belief that has already accumulated its own evidence of use. That is precisely the situation the correction stage has to deal with on its own terms.

Studying it

This corresponds to the algorithm folk theory paradigm commonly used in recent platform-algorithm research: users spend an extended period using an opaque recommendation or ranking system without ever getting an explanation of its internal logic. Researchers interview users to collect their own accounts of "how this algorithm works," compare those accounts against the system's actual logic, and record how users adjust their behavior based on the folk theory (for example, deliberately performing certain actions to "please the algorithm").

Common independent variables: how opaque the system is, how long the user has used it, how often the user encounters anomalous or surprising results. Common dependent variables: how closely the folk theory matches the actual mechanism, the type and frequency of behavioral adjustments made based on the folk theory, how stable the folk theory is (whether it persists even after encountering a counterexample).

Methodology note: folk theories are usually captured through interviews conducted after users have already used a product for quite some time, so what researchers get is an already-hardened version that's been through multiple rounds of self-confirmation — it's hard to reconstruct what it looked like right when the explanation gap first opened, before it hardened. This is also why post-hoc interviews alone can't easily settle the counterfactual question of "would this theory never have formed if an explanation had been given in time" — that counterfactual usually needs to be tested with an upfront intervention experiment, not a retrospective interview.

Where it stops holding

• Forming a folk theory requires repeated exposure to the same kind of unexplained anomaly — a single, isolated explanation gap isn't enough to make a user invest effort in building a whole causal theory. Only an anomaly that recurs and actually affects the user's decisions is worth the effort of inventing an account for.
• A folk theory isn't necessarily wrong in its conclusions — it can occasionally land close to the real mechanism by chance. This entry is about the risk built into how it forms (no verified basis, self-confirmed purely through repeated use), not a claim that every folk theory's specific content is incorrect.
• This entry only covers how a gap turns into a self-made causal belief — it does not cover how to correct that belief once it has hardened. That is no longer something an explanation gap framing can resolve; it requires handling the already-entrenched belief on its own terms.

Applying it

• For features already known to be long-term opaque and to repeatedly produce results users find confusing (recommendation ranking, auto-categorization, content distribution), proactively check for a long-standing, unfilled explanation gap — don't wait until a widely circulated "here's how it really works" story has already taken hold in the user community before noticing.
• Once a feature is found to have already spawned a widely circulated account among users, first verify how closely that account matches the real mechanism, rather than assuming "users have mostly guessed right." Wherever the match is poor is both where an explanation is most needed now and where the strongest resistance to correction should be expected.
• How to check: periodically collect users' open-ended descriptions of "how do you think this works" for opaque features, and compare them against the product team's actual implementation. Finding several high-frequency, mutually similar accounts that don't match the real mechanism is a sign that a folk theory has already formed and stabilized — meaning a single supplementary explanation is likely no longer enough.

Related

• Same group: A7.11.1 The explanation gap is the absence of an attributable cause after a result occurs — not the surprise of the result itself · A7.11.2 Timely explanation and after-the-fact explanation differ in how well they repair the model — timeliness itself has value · A7.11.3 An explanation needs to land on concepts already in the user's model; a generic apology can't fill the gap
• Nearby: A7.12 Causal attribution and superstitious behavior · A7.14 Recognizing and correcting wrong mental models
• Search terms: folk theory · algorithm folk theory · explanation gap · belief perseverance