Confusion is a map, not a mess

Every founder we talk to says a version of the same sentence: “People just don’t get it.”

They’re usually right, and they usually mean it as a complaint, as if the audience were being lazy, or the market were too early, or the message needed to be louder. That reading throws away the most useful piece of data a project has before launch. “People don’t get it” is where the inquiry starts, not where it ends, because the not-getting has a shape.

Here is the claim this whole desk rests on. When people misread a project, the misreadings are not random noise. They land in a few particular places, they travel the same routes across different readers, and they repeat, because they’re produced by ordinary machinery of cognition that has been studied for a century. The reader’s mind is not malfunctioning when it lands on the wrong reading. It’s doing exactly what minds do with an ambiguous new thing, and what it does is regular enough to predict. Confusion is understanding that went somewhere specific, just not where you wanted it to go. And because the causes are structured, the confusion is structured too, which is the only reason any of this can be drawn.

People build a story before they have the facts

Start with what a mind does the instant it meets something it can’t place. It doesn’t wait for enough information. It reaches for the most plausible account it can assemble from what’s already in front of it and commits to that, because a coherent-enough story is what lets a person act.

Karl Weick spent a career describing this in organizations, and the load-bearing line is that sensemaking is driven by plausibility rather than accuracy: people settle on the account that hangs together and lets them move, not the one that’s correct (Weick, Sutcliffe & Obstfeld, 2005; the book-length treatment is Weick, 1995). The implication for a launch is unforgiving. A visitor who can’t tell what your project is will not hold the ambiguity open and wait for clarification. They will manufacture an answer on the spot, from the nearest available material, and then treat that answer as what your project is. The gap you left gets filled whether or not you fill it, and it gets filled with whatever was lying around: the category next door, the last thing that looked like this, the reputation of the thing it resembles.

That is the first reason confusion isn’t a fog. A fog is uniform. This is directional. The mind moves toward the specific plausible story, and if two visitors share the same nearby material, they move toward the same one.

They read the new thing through the old thing’s frame

Which nearby material? Whatever schema the reader already carries. Frederic Bartlett established in 1932 that memory and comprehension are reconstructive: we don’t record a new thing as it is, we assimilate it to the frameworks we already hold, and we reproduce it bent toward those frameworks (Bartlett, 1932). His subjects, told an unfamiliar folk tale, retold it with the strange parts smoothed into shapes their own culture recognized. That is how comprehension works: the unfamiliar gets pulled into the familiar, and different people holding the same familiar frame pull it the same way.

Schemas don’t just tint the reading. They generate detail that was never there. William Brewer and James Treyens sat people in an office, then asked what they’d seen. Subjects confidently “remembered” books and a filing cabinet that the room never contained, because an office schema says those things belong, so the mind supplied them (Brewer & Treyens, 1981). Read that as a warning about your launch. A visitor who slots your project into “another L2,” “another memecoin,” “another wrapper” doesn’t just apply the label. They import the whole package the label carries: its typical team, its typical scam, its typical failure, none of which you shipped. The frame manufactures a version of you and hands it to the reader as memory. This is why the same wrong reading recurs across strangers who never spoke to each other. No one is copying anyone. Each reader runs the nearest schema, and the schema does the copying.

Newcomers sort by surface, and they sort the same way

There’s a predictable structure to whose misreadings look like what, and it turns on expertise. Michelene Chi, Paul Feltovich and Robert Glaser gave physics problems to experts and to novices and asked each group to sort them. Novices grouped by surface features, the pulley in the picture, the inclined plane, whatever was visible on the page. Experts grouped by the deep principle needed to solve them, which the surface hides (Chi, Feltovich & Glaser, 1981). The finding generalizes far past physics: the less someone knows a domain, the more they classify by what shows on the outside, and because the outside is shared, their errors converge. A newcomer to your category will read your surface, your palette, your token name, your homepage’s first impression, and file you next to whatever it superficially resembles. Every newcomer reading the same surface files you in the same wrong drawer.

Eleanor Rosch explains why the drawer is the one it is. Categories organize around prototypes, a best example the mind treats as the center of the category, and a new thing gets judged by how much it resembles that prototype and pulled toward the nearest one (Rosch, 1978). An unfamiliar project doesn’t get its own fresh category. It gets absorbed into the closest existing one and inherits its meaning. So you can often predict the target of the misreading, not just its kind: you can name, in advance, which prototype your project will be mistaken for, because it’s the nearest one on the surface features a newcomer can see.

The smooth version wins, then repetition makes it feel true

Now the reading spreads, and it spreads on ease, not accuracy. Rolf Reber and Norbert Schwarz showed that when a statement is easier to process, people rate it as more true. The ease of taking it in gets misread as evidence of its truth (Reber & Schwarz, 1999). Adam Alter and Daniel Oppenheimer traced how many different kinds of ease, visual, linguistic, how readily it comes to mind, all feed one and the same metacognitive signal that the mind then treats as a cue to believe (Alter & Oppenheimer, 2009). The practical consequence is bleak for a project with a complicated true story. A clean, wrong, easy-to-repeat account of you will outcompete an accurate account that takes three careful sentences, because the clean one processes more smoothly and smoothness reads as truth. The misreading wins on fluency, not evidence, and it wins against you specifically because your real story is harder to say.

Then repetition finishes the job. Lynn Hasher, David Goldstein and Thomas Toppino showed in 1977 that people judge a statement more true simply for having encountered it before, whether or not it’s actually true, the illusory-truth effect (Hasher, Goldstein & Toppino, 1977). It isn’t a fluke of one study: a meta-analysis by Alice Dechêne and colleagues across dozens of experiments confirms repetition reliably raises perceived truth (Dechêne, Stahl, Hansen & Wänke, 2010). So a misreading that gets repeated, one commenter to the next, one thread to the next, hardens into something the audience experiences as an established fact about you, on repetition alone. This is the mechanism behind the thing founders find maddening: the wrong take only has to be said enough times to feel true, while the right one has to be argued every single time.

Once it’s believed, it resists the correction

The last reason misreadings are worth mapping is that they don’t wash out when you correct them. Hollyn Johnson and Colleen Seifert established the continued-influence effect: a piece of misinformation keeps steering people’s inferences even after they’ve seen, accepted, and remembered a clear retraction (Johnson & Seifert, 1994). The false account is already woven into the mental model, and the “this was wrong” tag is weaker than the belief it was meant to cancel. Stephan Lewandowsky and colleagues generalized the finding and, usefully for a method note, catalogued what actually reduces it, which is far more than a single denial (Lewandowsky et al., 2012). And once a reader has landed on an interpretation, Raymond Nickerson’s review of confirmation bias says they’ll preferentially notice and weight whatever fits it, deepening the initial misread with each new piece of evidence they filter through it (Nickerson, 1998).

Stack those two and you get the property that makes early misreadings so expensive: they’re self-sealing. The first wrong frame doesn’t just persist through correction, it recruits confirming evidence and screens out the rest, so it grows more entrenched the longer it sits. A misreading caught on launch day is a task. The same misreading six months later is a fortification.

Structured causes, structured confusion

Read the chain back and the argument is complete. People impose a plausible story on ambiguity before they have the facts (sensemaking). They build it from the schema the nearest familiar thing left behind, which even generates false detail (schema theory). Newcomers do this by surface features, so their errors converge on the same prototype (categorization). The smooth version outruns the accurate one and repetition sets it as fact (fluency, illusory truth). Then it resists correction and recruits confirmation (continued influence, confirmation bias). Every step in that sequence is structured. Structured causes can’t produce random effects. So the misreadings arrive structured, which is the whole permission for the method: a structured thing has a shape, and a shape can be drawn.

Confusion isn’t the absence of a message. It’s a map of where yours breaks.

The moment you draw the map, the problem changes character. “People don’t get it” is a mood. It has no edges, so there’s nothing to do with it except worry or shout. But “here are the six places a skeptical visitor gets lost, ranked by how much each one costs you at launch” is a task list, and you can work a task list. That’s the move this desk makes, over and over: take the vague dread that a project is being misunderstood, and convert it into a diagram, the project in the middle, its likely misreadings clustered around it, each one sized by risk and colored by how sure we are it’s happening.

FIGURE · IN PRODUCTION

A confusion-map detail: a central project node with its clustered misreadings around it, each dot colored by evidence state and sized by launch cost, with the cognitive mechanism that produces it labeled on the edge. The full map anchors the trust audit case study.

What the science licenses, and what it doesn’t

One honest boundary, because the strongest version of this claim overreaches the evidence. Everything above establishes that misreadings are structured and non-random. It does not establish that they always collapse to a small, tidy, rankable handful, or that they follow any neat numeric distribution. That tighter claim, that in practice the misreads keep landing in five or six of the same spots, is the desk’s own wager from doing the work, not a result you can cite from cognitive science. In the launches we’ve mapped, the concentration holds: a name gets confused with one particular other name, a fork inherits one particular parent’s reputation, a claim raises one particular doubt in one particular audience. But that’s an observation from our own casework, offered as such, not a law of the mind. The science earns “mappable.” The rankable short list is what we keep finding when we actually draw the map.

The nearest prior version of “map the misunderstanding, then fix the top few” already lives in usability work, heuristic evaluation, card-sorting, mental-model mismatch testing. What’s added here is narrower: treating a project’s misreadings as a cognitively-structured object with named mechanisms behind each cluster, for a brand and product launch, rather than as feedback to be waved away. The map isn’t a new discovery. Naming why each dot on it is there, and predictable, is the part worth having.

Why this beats being louder

The instinct, when people misread you, is to say the true thing again, bigger and more often with more budget behind it. It rarely works, and now you can see the mechanism it’s fighting. You’re competing with a reading that already formed on plausibility, that fits an existing schema, that’s smoother than your correction, that repetition has already made feel true, and that will resist the retraction and recruit evidence against it. Volume doesn’t reach any of that. A louder true claim is still the harder-to-process one, so fluency keeps favoring the misreading, and the continued-influence effect keeps the misreading running after your correction lands.

Confusion has to be removed at the exact spot where it forms. You find the specific conflation, this name with that name, this fork with that scandal, and you interrupt it there, at the surface feature or the schema trigger that produces it. That’s cheaper than a louder launch and it holds, because you changed the mechanism instead of talking over its output.

Which is why the desk exists, and it fits in one line: if you can draw it, you can fix it. A project that knows precisely where it’s misread, in what order, at what cost, and by which mechanism, is in a completely different position from one that just feels misunderstood. The first has a plan. The second has an anxiety. Turning the second into the first is most of the job, and it starts by refusing to treat confusion as a mess.

It’s a map. Read it.