I. Before You Say a Word

There is a moment most people have experienced but few have named. You walk into a room — a classroom, a meeting, a disciplinary hearing, a doctor’s office, a job interview — and you can feel, before a single word is spoken, whether you will be believed.

It is not always visible as a single event. Sometimes it accumulates slowly, across many rooms, many years, many institutions. You notice that some people seem to arrive with a kind of authority already attached to them. They speak and the room listens. Their accounts of events are received as plausible. Their distress is treated as proportionate. When conflict arises, the default assumption tilts in their favor.

Other people walk in carrying the opposite. They have to establish credibility from scratch, in real time, under conditions that are already working against them. Their accounts are questioned before they finish speaking. Their distress reads as instability. When conflict arises, the default assumption tilts against them — not because of anything they’ve done, but because of how they are read.

This is not paranoia. It is not a distortion of experience. It is a feature of how institutions allocate legitimacy — and it operates, largely, before any assessment of what a person has actually said or done.

Some people arrive with credibility already assigned. Others spend their entire lives trying to earn what was never going to be given.

For many neurodivergent people, this experience is not occasional. It is the background condition of institutional life. The same dynamic — speaking clearly, being heard as confused; reporting accurately, being treated as unreliable; asking reasonable questions, being read as difficult — repeats across different settings with different people, generating the same outcome so consistently that it begins to feel like a personal failing.

It is not a personal failing. It is a measurement problem. And understanding it as a measurement problem changes everything about what the solution looks like.

II. What Credibility Is Actually Measuring

Credibility is the currency of institutional life. It determines who is listened to, who is trusted, who is believed when accounts conflict. And institutions talk about credibility as though it were a neutral, accurate measure of truthfulness and reliability.

It is not.

Credibility, as it functions in most institutional settings, is a social judgment — assembled from cues that correlate with social fluency far more reliably than they correlate with accuracy. Calm delivery. Confident eye contact. Linear, coherent narrative. Emotional responses that match what the evaluating institution expects. The ability to mirror the communication style of the person assessing you. These are the signals institutions read as evidence that someone is credible.

None of them measure honesty. They measure performance. And the performance that reads as credible was calibrated around a specific kind of communicator — one whose neurological and social profile happens to match the assumptions built into the evaluation.

Credibility, as it functions in most institutions, does not measure truthfulness. It measures how closely your presentation matches the institution’s template for a trustworthy person.

The consequences of this miscalibration are not evenly distributed. Neurodivergent people — whose genuine responses often include flat affect, atypical eye contact, intense or fragmented emotional expression, direct communication that reads as aggression, non-linear narrative — are read as less credible not because they are less honest, but because their honest responses don’t match the template.

Meanwhile, people who understand credibility signals and can perform them strategically — regardless of whether their account is accurate — receive the benefit of the doubt the system was always going to give to whoever performed most fluently.

The institution believes it is assessing truthfulness. It is assessing social performance. That gap between what is claimed and what is measured is where a great deal of injustice quietly lives.

III. The Proxies We Mistake for the Thing Itself

Credibility is not the only place this happens. Across nearly every domain where institutions claim to evaluate human capability, the same pattern appears: an institution chooses a proxy for something it wants to measure, the proxy is easier to observe than the thing itself, and over time the institution begins treating the proxy as though it were the real thing.

The gap between the proxy and the thing it was meant to represent does not disappear. It widens — quietly, systematically, in ways the institution rarely examines.

Education

Schools claim to measure learning, intelligence, and understanding. What they more often measure is compliance, tolerance for particular task formats, speed of information recall under time pressure, and the ability to sit still and attend to externally imposed structures for extended periods.

A student who deeply understands a concept but cannot demonstrate it under timed exam conditions is measured as less capable than one who has memorized information and can reproduce it quickly. The system calls this achievement. It may actually be measuring compatibility with the assessment format.

Hiring

Hiring claims to measure job ability. Most interviews are, functionally, measures of social performance under pressure — the candidate’s ability to mirror the interviewer’s communication style, produce confident and coherent self-narrative on demand, fill silence with the right kind of words, and display the non-verbal cues the interviewer reads as trustworthy.

A brilliant engineer and a mediocre engineer can receive opposite outcomes from the same interview process, because the interview is not measuring engineering. It is measuring interviewing. Those are different skills, and conflating them produces predictable errors at scale.

Performance Reviews

Performance reviews claim to measure contribution and output. Research on workplace evaluation consistently finds they more often measure visibility, self-promotion skill, political acumen, and manager affinity. The employee producing the most value is not always the employee receiving the best review, because the review process is not calibrated to capture value. It is calibrated to capture impression.

Professionalism

Professionalism is presented as a measure of competence and reliability. In practice, it is largely a set of conformity signals: eye contact norms, tone of voice, body language, small talk fluency, office etiquette. None of these necessarily correlate with ability to do the job. All of them influence evaluations of whether someone is capable of doing the job.

Framed differently: professionalism often measures how closely a person resembles the dominant social group’s definition of a capable person. It is not a neutral standard. It is an inherited one — and its costs fall unevenly on everyone who did not grow up inside the norms it encodes.

Leadership

Leadership selection claims to identify the most capable people to guide organizations. Research on leadership perception and advancement consistently finds that charisma and confidence are the traits most reliably rewarded — and that they are systematically mistaken for competence and effectiveness.

Some of the worst leaders look like leaders. Some of the best leaders don’t. The institution continues selecting for the appearance because the appearance is legible and the actual quality is difficult to see until the consequences arrive, by which point the selection has already been made.

Credibility

And here the circle closes. Credibility claims to measure truthfulness and reliability. It measures social legibility — the ease with which a person is read and processed by the institution evaluating them. The person who fits the template is believed. The person who does not has to argue for standing they were never going to be given on the same terms.

The pattern is consistent across every institution that claims to evaluate human capability: the proxy is not the thing. And the proxy was not chosen neutrally.

IV. The Consequences Are Not Limited to Neurodivergent People

This is where the argument opens into something much larger.

If a measurement system is measuring the wrong thing, the consequences are not limited to the people most obviously excluded by it. They radiate outward — to the institutions doing the measuring, to the teams and communities inheriting the outcomes, and eventually to the broader society those institutions are meant to serve.

For Neurodivergent People

The most visible consequence is exclusion. Talented people are filtered out of hiring processes that measure interview performance instead of job capability. Capable students are labeled deficient by schools measuring compliance instead of understanding. People spend years masking — performing the social signals that grant institutional access — rather than contributing from a position of genuine fit.

The individual cost is unemployment, underemployment, burnout, misdiagnosis, and a persistent erosion of confidence in one’s own perception and capability. These are not small costs. They accumulate across a lifetime.

For Institutions

Institutions believe they are selecting the best people. If the measurement is flawed, they are selecting the people who perform best on the measurement — which is a different thing entirely.

The consequence is systematic talent misidentification: missed innovators, specialists whose expertise doesn’t survive the interview, deep thinkers who don’t self-promote effectively, pattern recognizers who cannot perform the social rituals that signal belonging. Over time, institutions optimized for proxy performance become populated by people who are good at looking competent. That is not the same as being competent, and the difference eventually shows up in the outcomes.

For Teams

Teams inherit the consequences of poor measurement at every level. An organization that consistently promotes confidence over competence, visibility over contribution, and political skill over expertise does not simply promote the wrong individuals. It creates an environment where the most capable people — the ones least likely to self-promote, least likely to perform belonging, most likely to find the political dynamics alienating — eventually leave.

The remaining team absorbs the cost: lower capability, higher workload, greater dependence on people selected for the wrong reasons. The organization rarely traces this back to the measurement system that created it.

For Society

Every institution functions as a filter. Schools filter students. Hiring processes filter workers. Licensing systems filter professionals. Promotion structures filter leaders. If those filters are systematically measuring the wrong things, the aggregate effect is a society that misallocates human potential at scale.

The result is talent shortages in fields that claim to want talent but filter for performance. Leadership failures in organizations that selected for charisma rather than capability. Innovation loss from the systematic exclusion of thinkers whose cognitive profiles don’t translate into institutional legibility. These are not hypothetical costs. They are measurable, and they compound.

For Truth

This consequence is the most consequential of all.

If institutions systematically mistake confidence for competence and social fluency for truthfulness, then over time they begin rewarding people who are skilled at appearing right rather than being right. That changes who gets listened to, who gets promoted, who gets funded, who gets believed when it matters.

A system optimized for appearances does not simply make mistakes. It becomes unable to recognize its mistakes — because the measurement keeps producing the results the system expects.

The most capable people look like evidence that the selection was wrong. The selected people look like proof that the system chose correctly. The institution never examines what it was actually measuring, because the outcomes appear to confirm the method.

This is the deepest design failure: not that flawed measurement produces bad outcomes, but that it produces bad outcomes while convincing everyone the outcomes are correct. The excluded people become evidence that the system is working. The system becomes unfalsifiable from inside itself.

V. This Is a Design Problem

The gap between what institutions claim to measure and what they actually measure was not designed with malicious intent. It emerged from the natural human tendency to rely on observable proxies for things that are harder to see — and from the failure to examine those proxies against the outcomes they were meant to predict.

But the fact that the design failure was not intentional does not make it neutral. Unexamined systems do not sit still. They calcify. The proxies become the standard. The standard becomes the definition. And eventually, the institution loses the ability to distinguish between the signal it was trying to read and the noise it has been reading all along.

Understanding this as a design problem rather than a bias problem changes what the solution looks like. Bias interventions ask people to see past their assumptions. Design interventions change the measurement instrument. The first requires individual effort, conversation by conversation, assessment by assessment, against a current that is always running the other way. The second changes the current.

That means making the assumptions explicit. Writing down what a process claims to measure. Auditing whether the observable signals being used as proxies actually correlate with the claimed measure. Asking who is systematically excluded by the current instrument and whether that exclusion is evidence of a population problem or a measurement problem.

In most cases, when the audit is done honestly, the answer is the same: the measurement instrument was not designed for the full range of human capability. It was designed, largely without awareness, around a narrower model — one that treated a particular kind of social performance as universal, and read everything that diverged from it as deficiency.

The people excluded by a measurement system are not evidence that they failed. They are evidence that the measurement was not designed for them.

This reframe matters enormously for the individuals who have spent years inside the standard explanation — the one that locates the problem in them. The legitimacy gap is not something that can be closed by becoming more fluent in the signals the institution rewards. It can be closed by redesigning the instruments that mistake those signals for the thing itself.

That is a harder task. It requires institutions to examine something they have largely taken for granted: the assumption that their methods of evaluation are measuring what they say they are measuring.

Some institutions are beginning to ask that question. More need to.

Conclusion: What the System Was Built to See

The legitimacy gap is not a new problem. It is an old design running without examination. Institutions inherited assumptions about what trustworthy, capable, intelligent, and reliable people look like — and built their measurement systems around those assumptions so thoroughly that the assumptions became invisible.

The result is a set of filters that claim to be selecting for capability while selecting, in practice, for a narrow set of social performances that correlate with capability only weakly, under conditions that were never examined, in populations that were never representative.

For neurodivergent people, the gap is most visible and most costly. But the gap does not affect only neurodivergent people. It affects everyone whose genuine capability does not translate cleanly into the proxy the institution has decided to measure. It affects the institutions doing the measuring, which believe they are selecting the best and are often doing something else. It affects the societies those institutions serve, which are absorbing the compounding cost of misallocated human potential without being able to trace it to its source.

The question worth asking is not why certain people keep failing institutional measurements. That question accepts the measurement as accurate and looks for the problem in the person.

The question worth asking is what the measurement was actually built to see — and whether what it was built to see has anything to do with what the institution claims to value.

Most institutions are not failing at inclusion. They are succeeding at something else — a narrower, older task they inherited without examining. The work is not to make the excluded more legible to the system. The work is to build systems that can see.

Nothing is neutral. Normal was designed.

The people excluded by a system are not evidence that they failed. They are evidence that the design did.

© 2026 Eva Redford

This essay is part of an ongoing body of work examining how institutions are designed, who they are designed for, and what it would mean to build them differently. The argument here is not abstract. It is a blueprint.

Imagine you are applying for a job. Not a bad job. A good one. The kind you are qualified for. You have done the work, developed the skills, and built the knowledge the role requires. You are prepared.

Then comes the interview.

You are placed in a room with a stranger. You have thirty minutes. You are expected to perform confidence, maintain eye contact, modulate your voice for warmth, fill silences, answer unpredictable questions in real time without notes, and simultaneously demonstrate analytical ability, cultural fit, and enthusiasm for a company you have had forty-eight hours to research. The person across from you has been trained to evaluate you based on whether your performance matches a particular script for how a competent, motivated person looks.

If you are neurologically similar to the person who designed that script, this may feel difficult but manageable. If you are not, it may feel impossible. Not because you are less qualified. Because the system was not built to read you.

Now imagine the alternative.

You submit a work sample. You complete a short structured task. You have a conversation with a future colleague about a problem the team is actually working on. There is no performance. There is no script. There is only evidence of what you can actually do.

In the second scenario, the institution learns something real. In the first, it learns only who is best at performing competence for an audience. Those are not the same thing, and institutions that cannot tell the difference will keep selecting for the wrong signal.

Most institutions believe they are measuring capability.

In reality, many are measuring social legibility.

The result is not merely exclusion.

It is measurement error.

This is what redesign looks like. Not accommodation. Not inclusion programming. Not a revised mission statement. A different question asked at the beginning:

What does this system actually need to learn about a person, and what is the most accurate way to learn it?

Most institutions have never asked that question. They inherited their processes, optimized them for efficiency, and called the result merit.

This essay is about what happens when we ask it.

I. Why Reform Has Failed So Far

Before describing what redesign looks like, it is worth being precise about why existing reform efforts have not worked. This matters because the two are easy to confuse. Organizations that have added accessibility offices, developed diversity initiatives, and published inclusion commitments have not, in most cases, changed how their systems function. They have changed how their systems talk about themselves.

The distinction is not cynical. It is structural.

Reform efforts tend to operate at the level of exception. They identify people who do not fit the existing design and create pathways to accommodate them without altering the design itself. Disability accommodations. Extended time. Flexible dress codes. Quiet rooms. These interventions acknowledge that the default is not working for everyone. But they leave the default intact. They say:

Here is a workaround for people our system was not built for.

Redesign says something different. It asks why the default was built this way, who benefits from keeping it, and what would be lost if we changed it. In most cases, the honest answer is: very little of substance, and a great deal of social convenience.

The reason systems are not redesigned is rarely that redesign is technically impossible. It is that the current design works well for the people who have the power to change it. The interview process sorts effectively for candidates who look like the people already inside the organization. The classroom rewards the participation style of students whose home environments prepared them for that specific communication format. The performance review system advances employees whose output is highly visible, highly verbal, and easily legible to managers who do not look closely.

None of this is intentional exclusion in the way people usually mean that phrase. It is structural. The system produces predictable outputs because it was built on predictable assumptions, and those assumptions have never been seriously examined.

This distinction matters more than it might seem. When exclusion requires intent to exist, institutions can defend themselves by pointing to the absence of malice. No one meant to exclude. No one sat down and decided this system would disadvantage neurodivergent people. That may be true, and it changes nothing about what the system produces. Most exclusion is not the result of prejudice. It is the result of routine. Practices designed for a narrow range of people, repeated without examination, embedded into policy and culture, and eventually mistaken for neutrality. The harm is real regardless of the intent behind it, which means the solution cannot be located in intent either. You cannot fix a design problem by assuming good faith.

This is the first thing that has to change: institutions must become willing to examine their own defaults.

There is a sharper version of this problem that rarely gets named directly. Most inclusion initiatives are not designed around access for neurodivergent people. They are designed around comfort for the majority. The quiet room is there so that neurotypical colleagues do not have to think about sensory needs. The diversity statement is there so that leadership feels they have addressed something. The accessibility coordinator exists so that everyone else can feel absolved. These are comfort measures. They manage the discomfort of the majority with the idea of difference. They do not meaningfully change what a neurodivergent person encounters when they try to move through the institution.

Access and comfort are not the same thing. Access means the system can actually be used. Comfort means the people already inside it do not have to feel bad. Reform that optimizes for the second while calling it the first is not reform. It is reputation management.

III. What Redesigned Education Would Look Like

Education is the clearest case because the stakes are highest and the gap between stated and actual purpose is most visible. Schools say they exist to develop the capacities of every child. The way most schools are designed, they exist to develop the capacities of children who process information quickly, sit still, express themselves verbally on demand, and perform competence in front of an audience. For children who do not match that profile, school is less an education and more a long series of signals that something is wrong with them.

Redesigned education does not start with the question: how do we support children who struggle? It starts with: why are so many children struggling in the same ways, and what does that tell us about our design?

Instruction becomes multi-modal by default

In most classrooms, the primary mode of instruction is verbal delivery to a group. Students are expected to process spoken information in real time, retain it, and demonstrate retention through written output or verbal performance. For students who are auditory processors, this works. For students who need to read, see, move, or interact with information to retain it, this is already a mismatch before the lesson has begun.

Redesigned instruction does not assign alternative formats to students who request them. It builds multiple formats into every lesson as standard. Recorded audio. Visual diagrams. Written summaries. Time to process before responding. These are not accommodations for the few. They are better design for everyone.

Demonstration of learning is flexible

The current standard for demonstrating learning is overwhelmingly narrow: timed written exams, structured essays, verbal presentations, multiple choice questions. These formats measure, in large part, how comfortable a student is with the format itself. A student who understands the material deeply but cannot produce it on demand under timed conditions is indistinguishable, in most systems, from a student who does not understand it at all.

Redesigned assessment assumes that the same understanding can be demonstrated in multiple ways, and that the institution’s job is to find out what a student knows, not to filter for students who can perform under a specific set of conditions. Portfolio assessment. Project-based demonstration. Oral defense. Extended time as standard. None of these compromise rigor. They increase accuracy.

Struggle triggers a design question

This is perhaps the deepest change. In the current model, when a student struggles persistently, the system moves toward diagnosis, intervention, and remediation. The student is identified as having a deficit. Resources are directed at fixing the student.

In a redesigned system, persistent struggle first triggers a design question: is the instruction reaching this student, and if not, why not? This does not mean there are no students with significant learning differences that require specialized support. It means that the first response to struggle is not pathology. It is curiosity about the fit between design and learner.

The question shifts from: what is wrong with this child, to: what is our system failing to do for this child?

That is not a small change. It is a complete reorientation of where responsibility lives.

III. What Redesigned Hiring and Work Would Look Like

Employment is where design mismatch becomes most economically consequential. The hiring process functions, in practice, as a filter for a narrow social and neurological profile. It does this not because organizations intend to exclude but because they have inherited evaluation methods that were never audited for what they actually measure.

The job interview, as it is typically structured, measures: comfort with social performance under observation, verbal fluency under pressure, the ability to anticipate what an interviewer wants and deliver it, and familiarity with professional norms that are learned, not innate. It measures these things reliably. What it does not reliably measure is the ability to do the job.

Evaluation becomes evidence-based

Redesigned hiring starts by asking what the job actually requires. Then it designs evaluation to measure those specific things directly.

For a role requiring written analysis, the evaluation includes a written analysis task. For a role requiring collaborative problem-solving, the evaluation includes a structured collaborative problem. For a role requiring technical judgment, the evaluation includes a technical judgment task. The interview, if it exists at all, becomes a conversation about the work, not a performance of fit.

This approach has been piloted by enough organizations to have evidence behind it. Work sample tests are among the most predictive hiring methods known. They have lower adverse impact across race, gender, and neurotype than unstructured interviews. They are more accurate, more equitable, and, for most candidates, less alienating. The reason they are not universal is not that they are hard to implement. It is that they require organizations to be precise about what the job actually is, and many organizations have never done that work.

Communication expectations become explicit

One of the most significant sources of design mismatch in workplaces is the gap between stated expectations and actual expectations. Organizations publish job descriptions that list technical requirements. They then evaluate, promote, and retain employees based on social expectations that were never written down: how someone presents in a meeting, whether they ask questions in public or in private, how visible their work is to people with power, whether they seem enthusiastic in the right way at the right moments.

In a redesigned workplace, these expectations are surfaced and examined. Some of them describe real requirements. A client-facing role genuinely requires specific communication capacity. Many of them do not describe requirements at all. They describe the comfort preferences of the people already inside the organization.

Separating these two categories is not complicated. It is uncomfortable, because it requires people in positions of power to acknowledge that some of what they call professional standards are simply preferences, and that enforcing preferences as standards costs the organization real talent.

This is where inclusion initiatives most visibly fail neurodivergent workers. An organization can run sensitivity training, publish neurodiversity hiring targets, and celebrate awareness months while simultaneously maintaining a meeting culture, a promotion system, and a communication norm that makes it structurally harder for neurodivergent employees to be seen as competent. The initiatives are real. The barriers are also real. They coexist because the initiatives were designed to make the majority more comfortable with the idea of neurodivergence, not to remove the actual friction neurodivergent people encounter. Comfort and access are different outcomes, and only one of them requires changing the design.

The physical environment is redesigned, not patched

Most workplace accommodations for sensory and neurological needs are retrofitted: a quiet room added after the open-plan office proved unworkable, a standing desk granted after a formal request, noise-canceling headphones purchased individually because the environment produces too much noise to concentrate in. These are patches on a design that was never built with variation in mind.

Redesigned workplaces build environmental options in from the start. Multiple zone types — focused individual work areas, collaborative spaces, low-stimulation rooms, areas that permit movement — give people genuine choice about where to do what kind of work. Adjustable lighting, acoustic paneling, and furniture that signals its purpose rather than defaulting to uniform rows of desks are not expensive innovations. They are decisions made at the design stage rather than the remediation stage. Organizations that have built this way report higher sustained concentration, lower noise-related stress, and notably, no objection from neurotypical employees who also, it turns out, do not enjoy working in loud, poorly lit, inflexible spaces.

Productivity is measured through output, not visibility

The default metric for performance in most organizations is some version of perceived engagement: how present someone seems, how often they contribute in meetings, how available they are, how enthusiastically they perform investment in the work. These are all measures of visibility. They are not measures of output.

Redesigned performance systems measure what the role was meant to produce. Quality of work. Impact of decisions. Reliability of delivery. Depth of thinking. These can be measured. They are less immediately legible than visibility, which is why organizations default to visibility. But visibility is not value, and systems that confuse the two will systematically undervalue their least performative contributors, who are often their most rigorous ones.

Career progression does not require self-promotion

Perhaps the most consequential design failure in workplace systems is that career advancement is not primarily a function of contribution. It is primarily a function of social legibility. People advance when the right people notice them, when they successfully advocate for their own visibility, when they are comfortable asking for things and comfortable in the social dynamics of institutional power.

None of these capacities are equally distributed across the human population. They are, however, learnable, which means that organizations effectively sort for who has had the most practice navigating social hierarchies. This is not a measure of professional capability. It is a measure of prior exposure to contexts that teach it.

Redesigned progression systems build in regular, structured review. They make criteria explicit in advance. They require managers to account for the full range of a person’s contribution, not only the most visible parts. They do not eliminate advocacy; they ensure that advocacy is not the only path.

Capacity is treated as variable, not fixed

Most workplace systems are built on an implicit assumption: that a person’s capacity is roughly constant across days, weeks, and months, and that significant deviation from that constant is a performance problem. This assumption is false for most people and structurally harmful for people with chronic illness, fluctuating conditions, or neurological profiles that produce variable energy, focus, or tolerance for specific kinds of work.

A person with chronic fatigue, an autoimmune condition, a mental health condition that cycles, or simply a neurotype that produces intense focus on some days and very little on others is not an unreliable employee. They are an employee whose output pattern does not match the system’s expectations of how output should be distributed across time. Those are different problems with different solutions. The first is a capability question. The second is a design question.

Redesigned workplaces build in flexibility across time, not just across format. Deadlines structured around output rather than hours. Seasons of intensity and recovery recognized as a legitimate work pattern rather than inconsistency. The understanding that someone who produces exceptional work in concentrated bursts is not less productive than someone who produces average work steadily — they are simply producing differently, and a system that cannot read that difference is losing information.

IV. What Redesigned Metrics Would Look Like

Systems change when what they measure changes. This is both the most powerful and most neglected lever in institutional redesign.

Most institutions measure success by asking: who made it through? Who graduated, who was hired, who was promoted, who persisted? These are outcome metrics. They tell you what the current design produces. They do not tell you what the current design is failing to produce, because the people it is failing are no longer in the data.

Redesigned measurement asks the harder question:

Who exited, and where, and why?

Drop-off analysis. Attrition patterns. Points at which specific groups consistently disappear from the pipeline. These data tell you where the design is failing before the failure becomes invisible. An organization that tracks only its graduates does not know what it is losing. An organization that tracks who leaves, and when, and under what circumstances, is looking at its own design assumptions.

Absence-aware systems

The shift required here is from measuring presence to measuring absence. Who is not graduating from this program? Who is not passing this stage? Who was never present at all?

Most institutions do not have systems for tracking absence because absence is harder to see than presence. Building those systems requires intention. It requires deciding that the people who are not here matter as much as the people who are. For organizations accustomed to selecting the best-fitting candidates from a large pool, this represents a fundamental reorientation: the question is no longer who fits this design, but whose capability this design cannot read.

Environmental stress indicators

Burnout, disengagement, chronic accommodation requests, and high turnover in specific roles or demographics are not personnel problems. They are design signals. They indicate that the environment is producing predictable stress responses in a predictable population, which means the environment is the variable, not the individuals responding to it.

Redesigned institutions track these indicators systematically, analyze the patterns, and treat them as feedback about design rather than performance. This requires a cultural willingness to be wrong about the design. That willingness is rarer than it should be, but it exists, and organizations that have developed it consistently outperform those that have not.

V. The Language Question

Language is not peripheral to this work. It is structural. The words institutions use to describe their practices shape what those practices are allowed to become.

When a school says a student has a “processing deficit,” it has named a child as the problem. When a school says there is a “design mismatch between instruction format and how this student learns,” it has named a design failure. These two framings produce entirely different responses. The first leads to an intervention directed at the student. The second leads to a question about the design.

The language shift that redesign requires includes:

From “support needs” to “design requirements.”

Support needs frames variation as a deficit requiring compensation. Design requirements frames variation as information the system should be built to accommodate.

From “accommodation” to “default flexibility.”

Accommodation implies exception. Default flexibility implies that the range is the baseline.

From “struggles with” to “experiences mismatch with.”

Struggle is internal. Mismatch is relational. One locates the problem in the person. The other locates it in the interaction between person and design.

From “culture fit” to “contribution capacity.”

Culture fit is a preference disguised as a standard. Contribution capacity is the actual question.

These are not just semantic changes. They shift where institutions look for solutions.

VI. Where This Is Already Happening

Redesign is not purely theoretical. Versions of it exist, have been tested, and have produced results worth examining.

In hiring, a growing number of organizations, including several in the technology sector, have shifted to structured, work-sample-based hiring processes and reported both improved retention and reduced demographic sorting. The evidence that unstructured interviews are among the least predictive hiring tools has been in the research literature for decades. Organizations that have acted on it have generally not regretted it.

In education, project-based learning environments, Universal Design for Learning frameworks, and mastery-based progression systems have demonstrated that students who perform poorly under traditional assessment conditions frequently perform at or above average when given alternative demonstration formats. The question is not whether these students are capable. The question is whether the system was designed to see their capability.

In workplace design, distributed and asynchronous work structures, which became widespread during the pandemic and have persisted in many sectors, have shown that a significant portion of what was believed to require in-person, synchronous presence was actually a preference, not a requirement. The expansion of remote work has functioned, in practice, as an inadvertent redesign experiment, and the results have been instructive.

None of these examples represent complete redesign. They are partial, often contested, and frequently reversed when organizational culture reasserts existing norms. But they demonstrate that the changes described in this essay are not imaginary. They are choices. Institutions that have made different choices have gotten different results.

II. Why Institutions Keep Mistaking Familiarity for Competence

If unstructured interviews are weak predictors of job performance — and decades of research confirm that they are — why do organizations continue using them? The answer is not ignorance. Most HR professionals have access to the same research. The answer is that unstructured interviews feel accurate. They produce a strong subjective sense of having assessed something real. And the reason they feel accurate is that they reward the same traits the evaluator already possesses: verbal fluency, comfort with social performance, familiarity with professional scripts. People trust evaluation methods that surface people who seem like them. That trust is not evidence. It is bias wearing the costume of judgment.

This is the core mechanism behind institutional inertia. It is not that institutions are committed to bad design. It is that the people with authority to change the design experienced it as fair. The system selected them. From inside that experience, the system looks like it works. The people it did not select are not visible from that vantage point — they are simply absent, and absence reads as irrelevance rather than as a design failure.

There is also a legitimacy loop at work. The people a system certifies, hires, and promotes become the evidence that the system is accurate. A particular profile becomes the definition of qualified — not because it is the most capable profile available, but because the system has selected for it repeatedly and its own authority is treated as proof. The system validates its criteria by selecting people who match them, then points to those people as evidence the criteria are correct.

Breaking this loop requires introducing information the loop cannot generate itself. That information exists. It lives in the experience of the people the system consistently fails to read — the people whose capability did not surface through the standard process, not because the capability was absent, but because the process was not designed to find it. Neurodivergent people are not edge cases in institutional design. They are among the most precise diagnosticians of where institutional design fails. The people a system struggles to accommodate are often the people exposing where its assumptions are least accurate. That is not a burden. It is a form of institutional intelligence that most systems are currently leaving outside the room.

VII. The Economic Argument

The case for redesign is moral. It is also structural. It is also economic, and that argument deserves to be made directly, because institutions that will not move on moral grounds sometimes move on efficiency grounds, and the efficiency case here is substantial.

When a hiring process filters for social performance rather than job-relevant capability, it produces two kinds of error simultaneously. It selects some people who are good at interviewing but not at the job. And it screens out some people who are poor at interviewing but excellent at the job. Both errors cost the organization. The first shows up as underperformance and eventual attrition. The second shows up as talent that was never captured at all — which is harder to measure precisely because the person is not in the data, but not harder to understand. Every time a system screens out a capable person on the basis of an irrelevant signal, it pays a cost it will never see itemized on a budget line.

Attrition is more visible and still consistently underestimated. The cost of replacing an employee is typically estimated at between fifty and two hundred percent of their annual salary, depending on the role and seniority. Neurodivergent employees who leave organizations because the environment was not designed for sustained function — because the sensory conditions were unmanageable, because the communication norms were exhausting to perform indefinitely, because the performance review system consistently failed to register their actual contribution — represent a preventable attrition cost that organizations are currently absorbing without examining its source.

Burnout is a related and compounding cost. The cognitive and emotional labor required to mask, translate, and perform neurotypical norms in environments not designed for variation is significant and largely invisible to the institutions producing it. It does not show up as a line item. It shows up as reduced output, increased sick leave, disengagement, and eventually departure. Organizations that treat this as a personnel problem rather than an environmental one will keep paying it indefinitely.

There is also an innovation cost that is harder to quantify but important to name. Cognitive diversity — genuine variation in how people process information, identify patterns, weight risk, and generate solutions — is one of the more reliable predictors of organizational resilience and creative output. Hiring systems that filter for a narrow neurological and social profile produce teams that are, by design, less cognitively diverse than the population they are drawn from. The cost of that homogeneity is not always visible in quarterly outputs. It tends to become visible when organizations encounter problems that their existing thinking patterns cannot solve.

None of this requires treating redesign as charity. It requires treating it as what it is: a correction of measurement error that is currently costing institutions more than the redesign would.

VIII. What This Requires of Institutions

Redesign is not a program. It is a posture. It requires institutions to hold three commitments simultaneously.

First: willingness to examine defaults. Every system embeds assumptions. Redesign begins when institutions become willing to name their assumptions, ask whose interests they serve, and evaluate whether they are justified by what the system is actually trying to accomplish.

Second: tolerance for interpretive uncertainty. Redesigned systems have to be able to hold the possibility that they are misreading capability. That a person who did not perform well on their evaluation may nonetheless be capable. That a student who could not pass the standard test may nonetheless understand the material. This tolerance is not soft. It is rigorous. It asks institutions to be precise about what they actually know versus what they are inferring.

Third: accountability for absence. Institutions that are serious about redesign track who is not here. They ask why people left. They treat attrition, disengagement, and systematic underperformance in specific populations as information about their own design, not about those populations.

These commitments require leadership that is willing to be uncomfortable. That is not nothing. But it is also not unusual. Every significant improvement in institutional practice has required someone with institutional power to decide that accuracy matters more than convenience.

A fourth commitment, often omitted: design with, not for. Institutions that are serious about redesign involve neurodivergent people in the design process itself, not as focus group participants after decisions have been made, but as people with design authority over the systems that affect them.

This is not a gesture toward inclusion. It is an epistemological requirement. The lived experience of navigating a system that was not built for you produces a specific and highly accurate form of knowledge that cannot be replicated by researchers studying that experience from outside it. A person who has spent years finding workarounds for a poorly designed process knows exactly where the process fails, why it fails, and often what a better design would look like. That knowledge is institutional intelligence. It is currently sitting outside most institutions, untapped, because the people who hold it were not invited into the room where design decisions are made.

When neurodivergent people are brought into design processes as consultants or compliance checkboxes, the institution gets a narrow and often tokenized slice of that intelligence. When they are brought in as designers — with actual authority to propose, reject, and revise — the institution gets something much more valuable: a direct line to the failure points it cannot see from inside its own assumptions.

This also changes what neurodivergent people are understood to be within institutional logic. Not recipients of accommodation. Not edge cases requiring management. Sources of design intelligence that the institution needs in order to function accurately. That is a different relationship entirely, and it produces different outcomes — for the institution and for the people inside it.

IX. The Core Argument

The systems that govern how we learn, get hired, advance in our work, and access resources were built around a narrow conception of how humans function. That conception is not neutral. It is not scientific. It was produced by specific people, in specific historical moments, to serve specific purposes, and it has been inherited and optimized ever since without sufficient examination. Most modern institutions were optimized for standardization and administrative efficiency, not for the accurate evaluation of human variation.

Standardization solved the problem institutions were designed to solve. The problem is that it was never the same thing as understanding human capability.

The result is a set of institutions that are, in measurable and predictable ways, less accurate than they need to be. They are selecting for social performance instead of capability. They are labeling mismatch as deficit. They are losing people whose contributions are not legible to systems that were not designed to read them.

This is not primarily an equity argument, though the equity implications are significant. It is an accuracy argument. Institutions that cannot read the full range of human capability are operating on incomplete information. They are making worse decisions than they need to make. They are producing worse outcomes than they could produce.

It is also not primarily a legal argument. Anti-discrimination law is necessary. It is not sufficient. A system can be fully compliant with every non-discrimination statute on the books and still produce, reliably and predictably, the exclusion of neurodivergent people at every stage of its operation. Rights-based frameworks establish a floor. They do not build the structure. Redesign requires proactive institutional responsiveness — not the absence of discrimination, but the active, ongoing, accountable work of building systems that respond accurately to the people inside them.

Redesign is not charity. It is precision.

There is a principle in design that is worth stating plainly here: when you design for the edges, you improve the experience for everyone. Curb cuts were built for wheelchair users. They are used constantly by people with strollers, delivery carts, bicycles, and aging joints. Captions were built for deaf viewers. They are used by people watching in loud environments, people learning a second language, and people who simply process text faster than audio. Every design improvement made for the people a system was failing turns out to make the system more functional for the people it was already serving. Institutions that treat neurodivergent design requirements as edge cases are misunderstanding where design insight comes from. The edges are where you find out what the middle actually needs.

It is also worth being honest about what universal design cannot do. Building flexible systems by default will reach more people than rigid systems ever will. But there will always be people whose needs are specific enough, variable enough, or complex enough that universal design alone is insufficient. The goal of redesign is not to make individualized support unnecessary. It is to make it the exception rather than the rule, and to ensure that when individualized support is needed, it is available without the person having to prove their right to exist in the institution first.

The question is not whether institutions can afford to redesign. The question is whether they can afford the ongoing cost of designs that consistently misread human capability. That cost is paid in potential unrealized, in talent excluded, and in the accumulated signal loss of every person a system failed to see clearly.

The people those systems failed to see clearly are not outliers. They are the evidence that the design has a problem.

Nothing is neutral. Normal was designed.

The measure of a well-designed institution is not how smoothly it runs for the people it was built for. It is how accurately it reads and responds to the people it was not. That is the standard. Everything else is optimization within a flawed default.

The people excluded by a system are not evidence that they failed.

They are evidence that the design did.

When the same experiences are repeatedly dismissed across different people and contexts, that is not randomness. That is structure.

WORKS CITED

Berne, Patty. “Disability Justice — a working draft.” Sins Invalid, June 10, 2015. Archived at https://web.archive.org/web/20231026185203/https://www.sinsinvalid.org/blog/disability-justice-a-working-draft-by-patty-berne

CAST. Universal Design for Learning Guidelines, Version 2.2. CAST, 2018. https://udlguidelines.cast.org

Charlton, James I. Nothing About Us Without Us: Disability Oppression and Empowerment. University of California Press, 1998.

Disability & Philanthropy Forum. “What Is Disability Justice?” disabilityphilanthropy.org, 2022. https://disabilityphilanthropy.org/resource/what-is-disability-justice/

Gallup. State of the American Workplace. Gallup Press, 2017. Cited for voluntary turnover cost estimates ($1 trillion annually, 50–200% of salary per departure).

Meyer, Anne, David H. Rose, and David Gordon. Universal Design for Learning: Theory and Practice. CAST Professional Publishing, 2014.

Rose, David H., and Anne Meyer. Teaching Every Student in the Digital Age: Universal Design for Learning. ASCD, 2002.

Schmidt, Frank L., and John E. Hunter. “The Validity and Utility of Selection Methods in Personnel Psychology: Practical and Theoretical Implications of 85 Years of Research Findings.” Psychological Bulletin, vol. 124, no. 2, 1998, pp. 262–274. Cited for predictive validity of work sample tests (.54) and unstructured interview comparisons.

Schmidt, Frank L., In-Sue Oh, and Jonathan A. Shaffer. “The Validity and Utility of Selection Methods in Personnel Psychology: Practical and Theoretical Implications of 100 Years of Research Findings.” Fox School of Business Research Paper, 2016.

Sins Invalid. Skin, Tooth, and Bone: The Basis of Movement Is Our People — A Disability Justice Primer. 2nd ed. Sins Invalid, 2019.

Society for Human Resource Management (SHRM). “Retaining Talent: A Guide to Analyzing and Managing Employee Turnover.” SHRM Foundation, 2008. Cited for employee replacement cost estimates (50–200% of annual salary).

Taylor, Steven J., ed. “Institutions and the Promise of Change.” Center on Human Policy, Syracuse University / PRO-ED, 2004. worksupport.com/documents/proed_instituationalchange.pdf. Cited for the human variation model and systemic mismatch framing.

Most institutions believe fairness means consistency.

The same rules. The same process. The same expectations applied to everyone, without exception.

On paper, that sounds fair.

In practice, it often isn’t. Standardization and fairness are not the same thing. But the deeper problem is not that standardized processes create barriers. The deeper problem is what those processes do before anyone even encounters a barrier.

They define what competence looks like.

And once an institution has defined competence, it has also quietly defined who appears competent and who doesn’t.

Preferences dressed as standards

When an institution designs a process, it makes choices. Which format. Which timeline. Which inputs count. Which behaviors get rewarded. How communication should sound. How learning should look. How performance should be demonstrated.

Every one of those choices encodes a preference. A preference for one cognitive style over others. For one communication register over others. For one way of processing information, managing time, tolerating sensory environments, and signaling capability.

That preference does not stay a preference for long.

Once it is built into a process, it becomes a requirement. Once it is a requirement, it becomes a standard. Once it is a standard, it becomes the definition of normal. And once something is normal, questioning it stops feeling like institutional critique and starts feeling like a personal failure to meet a reasonable expectation.

This is how preferences become standards, and standards become invisible.

The institution does not experience its process as a set of choices. It experiences it as a neutral framework. The people who perform well inside it appear talented. The people who struggle inside it appear deficient. But the field was never level. It was designed by people, for a particular kind of person, and then called objective.

Institutions do not discover normal. They define it.

This is the most consequential thing institutions produce: not policy, not procedure, but normality itself.

They decide what communication looks like. What professionalism looks like. What intelligence looks like. What competence looks like. And the resulting standard is routinely mistaken for an objective description of reality rather than a design choice made by people operating within a particular culture and cognitive framework.

Neurodivergent people often make these assumptions visible, not because they are more affected than all other groups, but because they are among those most likely to diverge from the institutional prototype in ways the institution did not anticipate and did not design for. When the prototype assumes one way of processing language, managing attention, tolerating sensory input, reading social cues, and demonstrating knowledge, neurodivergent people encounter not just a difficult process but a process that was not built with their existence in mind.

That gap between the person and the prototype is regularly interpreted as a gap in the person. The design remains unexamined.

The chain runs in one direction: preference becomes standard, standard becomes normal, normal becomes the definition of competence. By the end of that chain, the original preference is invisible. It has become reality. And anyone who does not conform to it has become, in the institution’s accounting, deficient.

The legitimacy problem

A standardized process does more than evaluate performance.

It defines what performance looks like.

A job interview does not simply identify qualified candidates. It defines which communication styles appear professional. A classroom does not simply measure learning. It defines what learning should look like. A performance review does not simply assess contribution. It defines which behaviors count as contribution.

Once those definitions are institutionalized, the people who naturally fit them appear competent. Those who do not fit them appear deficient. The institution mistakes familiarity for merit.

The result is a self-reinforcing legitimacy system. The people who fit the standard succeed. Their success validates the standard. The standard becomes harder to question. Those who are excluded are treated as exceptions, as individual shortcomings, rather than as evidence of a design problem.

That is why standardization is so powerful. It does not just distribute opportunities. It distributes legitimacy. And legitimacy, once distributed, shapes how people are seen, not just professionally, but fundamentally. As capable or incapable. As credible or not credible. As qualified or not qualified.

The institution calls this a meritocracy. What it has actually built is a system that rewards people who match its prototype, and then uses their success as proof that the prototype was right.

Legibility and credibility

The legitimacy problem does not stop at access. It extends to whose account gets believed.

Once institutions define what competence looks like, they also, as a consequence, define what credibility looks like.

The employee who communicates in familiar ways is seen as professional. The student who learns in familiar ways is seen as intelligent. The candidate who interviews in familiar ways is seen as capable. These legibility assessments quietly shape something beyond evaluation. They shape how much weight an institution gives to what a person says.

When someone falls outside those expectations, their account becomes easier to dismiss. Their requests are questioned. Their performance is scrutinized more closely. Their explanations are treated as excuses. The same traits that reduce institutional legibility tend to reduce institutional credibility, not because the person is less reliable, but because they present differently than the prototype the institution learned to trust.

This produces a specific and damaging loop. The people most affected by a design problem are frequently the least believed when they identify it. An employee who diverges from the communication prototype raises a concern about the performance review criteria. The concern is heard through the same evaluative lens that produced the problem. The employee appears difficult, defensive, lacking self-awareness. The design remains unexamined.

This is not a failure of individual judgment. It is a structural feature of systems that have mistaken their own prototype for an objective standard. Once a standard is treated as neutral, challenges to it can only appear as the challenger’s problem.

Hiring: the social legibility test

The job interview is the most visible site of this pattern.

Most organizations require every candidate to complete the same process. Same questions. Same format. Same unspoken expectations about eye contact, verbal fluency, affect, social timing, and the performance of confidence in an unfamiliar room with strangers evaluating every word.

The process is standardized. It is also, quietly, a test of social legibility: the ability to present oneself in ways that read as professional, capable, and trustworthy to a particular kind of evaluator.

The problem is not that social legibility is being assessed. The problem is that it is being assessed instead of job performance, while being labeled as an evaluation of job performance.

“Culture fit” makes this most explicit. It functions as a hiring criterion while treating the existing culture as neutral, as if the dominant communication style, the expected social register, the informal norms about self-presentation are simply how things are, rather than a set of preferences that certain people learned to embody and others didn’t.

When a candidate is screened out for culture fit, the institution has not assessed whether they could do the job. It has assessed whether they resemble the people already there. Those are different measurements. Treating them as equivalent is not a neutral choice. It is a design decision that favors conformity to an existing prototype and calls that conformity competence.

Workplaces: the institutionalization of one cognitive style

The same pattern extends into employment itself.

Many workplaces have communication norms that are treated as professional standards while functioning, in practice, as one cognitive style elevated into a universal requirement.

Direct communication gets interpreted as rudeness. A flat affect gets read as disengagement. A preference for written over verbal communication gets framed as inflexibility. Small talk gets treated as a relationship-building competency. The ability to process verbal instructions in real time, without written reference materials, is treated as a basic professional expectation rather than one particular way of receiving information.

None of these norms are neutral. They reflect the preferences of the people who had the most power to define what professionalism means. Because those people experienced the norms as natural, the norms became invisible as preferences. They became the baseline. The definition of what a capable professional looks like.

Performance reviews make this concrete. Criteria like “executive presence,” “stakeholder management,” “communication style,” and “tone concerns” frequently measure conformity to social expectations rather than outcomes. An employee can produce excellent work and receive critical feedback because the way they delivered that work did not match the expected register.

This is not performance assessment. It is a legibility test, one that measures how closely a person’s natural style matches the institutional prototype. When that test produces a gap, the gap is attributed to the person. The prototype remains unexamined.

Education: measuring compliance, crediting it as intelligence

Schools offer perhaps the most consequential version of this problem, because the judgments produced there follow people for the rest of their lives.

The standardized test is presented as an objective measure of ability. It also measures processing speed, test-taking strategy, stress tolerance, and the capacity to perform under artificial time pressure. These are not the same thing as knowledge or intelligence. But because the test is uniform, because everyone takes the same one, its results are treated as if they measure something intrinsic to the student rather than something produced by the interaction between the student and the test design.

The same logic runs through grading systems, participation requirements, and the general expectation that learning should look a specific way: seated, quiet, regulated, compliant, making eye contact with the speaker, raising a hand before speaking. These are not descriptions of learning. They are descriptions of one particular performance of learning, the performance that was legible to the people who designed the system.

When a student cannot produce that performance, the system codes them as a problem. Unmotivated. Inattentive. Behind. The language of deficit is applied to the student rather than to the design that produced the assessment.

The longer-term effect is not just an unfair grade. It is an institutional judgment about intelligence and potential, a judgment that gets recorded, reported, and carried forward into every subsequent institution that student encounters. Standardized processes in education do not just measure. They produce the categories of capable and deficient. And they produce those categories using instruments built around one cognitive prototype, presented as universal truth.

The better question

Institutions often believe they are measuring competence.

In reality, many are measuring resemblance. Resemblance to the people who designed the process. Resemblance to the people who already succeeded within it. Resemblance to a prototype that became a standard and a standard that became “normal.”

Real fairness requires a different diagnostic question. Not: “Did everyone receive the same process?” But: “Did the process accurately measure what it claims to measure, for everyone it was applied to?”

Those are not the same question. The first defends the process by pointing to its consistency. The second interrogates whether the process was valid in the first place.

Consistency is not validity. A process applied equally to everyone still cannot measure competence if it was designed to measure conformity to a specific cognitive prototype and then labeled as something else. Institutions that take the second question seriously are forced to examine the assumptions inside their own designs. What does this process actually require? Who does it systematically favor? Is what it rewards genuinely related to the outcomes it claims to predict? Or has familiarity been mistaken for merit?

Those questions are uncomfortable. They implicate the people who built and currently benefit from the system. They suggest that success inside the institution is not purely a function of capability. It is also, in part, a function of how closely a person’s natural style matches the prototype the institution decided to reward.

The question is not whether everyone was treated the same.

The question is whether the institution ever recognized the difference between competence and conformity.

The most powerful forms of exclusion do not announce themselves as exclusion.

They present themselves as standards.

They call conformity competence.

They call familiarity merit.

And they keep distributing legitimacy to the people who were always going to fit, while describing the outcome as fair.

© 2026 Eva Redford

Editor’s note: This is the unifying statement of a framework I have been developing in pieces (design mismatch, social legibility, the translation tax, institutional legitimacy), gathered here into a single argument. It does not claim that autism is not a disability, or that autistic people lack real support needs; the difficulties are real. It claims that those difficulties have been mislocated. To call autism a design problem rather than a deficit problem is not to deny that it can be disabling. It is to relocate where the disability is produced: in the fit between a real mind and an environment built for a different one, rather than inside the person alone.

The claim is therefore larger and more structural: that the institutions governing a modern life were built around a narrow, historically recent model of the standard human being; that this model has become invisible; that institutions distribute legitimacy according to proximity to it; and that a great deal of what they record as individual deficiency is the predictable output of that design. Autism is the case that exposes the machinery, but the machinery is general. The argument is meant to be analytical rather than activist: a foundation to build on, not a verdict to applaud.

When a wheelchair user cannot enter a building, we no longer say the person failed to climb the stairs. We say the building has no ramp. The disability is real; its disabling effect is a property of the interaction between body and building, not of the body alone. Disability studies has called this the social model of disability for half a century, and where physical access is concerned, the culture has largely absorbed it.

Where minds are concerned, it has not. When an autistic person cannot pass a job interview, sit through a standardized classroom, or convince a jury they are telling the truth, we still locate the entire problem inside the person. The environment is treated as a neutral container in which an individual deficit becomes visible. This essay treats that assumption as the thing to be examined, and it leads to a question that sounds like it is about people but is actually about systems:

What if autism reveals not only differences in people, but limitations in the systems built to accommodate human variation?

The deficit model has a built-in answer, and the answer is no. On that model, autistic difficulty is a property of autistic people, and the systems they move through are neutral instruments that merely detect it. But every measuring instrument is built to a specification, and a specification is a choice about what counts. So the operative claim of this essay can be put in a single sentence:

The question is not whether autistic people have differences. The question is whether institutions have mistaken incompatibility with their own designs for evidence of individual deficiency.

Deficit and design are not two words for the same thing. They are competing answers to a single question: where is the problem located? The deficit model places it inside the person and treats the environment as a neutral given. The design model places it in the relationship between the person and a built environment that was never neutral to begin with. This essay argues that for a great deal of what we attribute to autism, the second answer is the more accurate one. Not that the difficulty is imaginary, but that it is, in large part, manufactured at the seam between a mind and a system designed without it in view.

The Deficit Inheritance

The clinical concept of autism is younger than the automobile. Leo Kanner, a child psychiatrist at Johns Hopkins, described eleven children in 1943 under the label “early infantile autism,” borrowing the word autistic from Eugen Bleuler, who had used it for a turning inward in schizophrenia (Kanner, 1943). A year later, working independently in Vienna, the pediatrician Hans Asperger described a partly overlapping group (Asperger, 1944). From the beginning the condition was named and defined from the outside, by what it looked like to the observing clinician, never by how it was experienced from within.

That outsider’s vantage point set the trajectory. Kanner did not stop at description; in noting that his patients’ parents were cold and intellectual, he seeded the “refrigerator mother” theory that would dominate explanation for decades before collapsing (Kanner, 1943). Autism began its institutional life as damage attributed first to the child and then to the mother. The diagnostic apparatus then formalized the framing. Victor Lotter’s 1966 study, the first of its kind, estimated prevalence at roughly 4.5 per 10,000: autism as rare and severe (Lotter, 1966). Lorna Wing and Judith Gould’s 1979 Camberwell study, and Wing’s later popularization of Asperger’s work, reconceived it as a spectrum defined by a “triad of impairments” (Wing & Gould, 1979; Wing, 1981). Infantile autism entered the DSM-III in 1980; Asperger’s disorder and PDD-NOS widened the category in 1994; the DSM-5 unified the subtypes into Autism Spectrum Disorder in 2013. Measured prevalence climbed from about 1 in 150 eight-year-olds in 2000 to 1 in 31 in the CDC’s most recent surveillance year, released in 2025 (CDC, 2025).

Every revision shares one feature, and it is the feature that opens the whole inquiry. The criteria are, without exception, a list of deficits, deviations, and restrictions, each measured against a standard the manual never states.

Deficit Requires a Reference Model

This is the hinge of the entire argument, and it is almost never examined: a deficit is not a freestanding fact. It is a comparison. To call something lacking is to measure it against a reference model of what a sufficient version would be, and the reference model is doing all the work. “Deficient social communication” means deficient relative to an assumed norm of communication. “Restricted interests” means restricted relative to an assumed normal range. Remove the reference model and the deficit dissolves into mere description: this person communicates in this way, attends to these things, moves like this.

So the deeper question is never “why is autism framed as a deficit?” The deeper question is “compared to what?” And the answer is almost always invisible, because the reference model is not written into the diagnosis. It sits behind it, unstated, doing the comparing while escaping the scrutiny. A diagnostic manual that lists what autistic people fail to do, without ever defining or defending the standard they are failing, has not described a pathology. It has described a distance from a prototype it declines to name. To evaluate the deficit, then, we have to do the one thing the deficit model never does: turn around and examine the standard itself.

Normal Was Designed

The standard has a history, and it is shorter than people assume. Normal, in the evaluative sense we now use it, is a nineteenth-century invention. As the literary scholar Lennard Davis has documented, until roughly the 1840s the English word normal meant “perpendicular”; it came from the carpenter’s square, the norma, and described a right angle, not a kind of person (Davis, 1995). For most of human history there was no concept of being normal to fail at.

The concept arrived with statistics. The Belgian statistician Adolphe Quetelet, in his 1835 treatise on man, introduced l’homme moyen, the average man, and made a quiet but consequential move: he treated the statistical average not as a description of a population but as an ideal, with deviation recast as a kind of error (Quetelet, 1835; Davis, 1995). Francis Galton, who coined the word eugenics in 1883, took the resulting bell curve and ranked it, declaring some deviations superior and others inferior (Davis, 1995). Within a few decades, a tool invented to describe groups had become an instrument for sorting individuals into the desirable and the defective.

This matters because the institutions of modern life were built in the same period and on the same logic. The factory required interchangeable workers on synchronized time. Mass schooling, designed in part to supply them, adopted the same architecture: age cohorts, fixed schedules, uniform instruction, every child measured against one developmental track. Competence was redefined as conformity to the track.

The crucial point is that these systems were not built to accommodate variation. They were optimized for sameness, because standardization and predictability were the point; interchangeability was the design goal, not an accident of it. Human variation was treated as friction to be engineered out, and a mind that could not be made interchangeable registered as a defect in the worker rather than a limit of the design. Normal, in other words, was not found in nature and then built into institutions. It was manufactured as a statistical artifact, elevated into an ideal, and engineered into systems optimized for uniformity, which then mistook their own optimization target for the natural shape of a competent human being. A baseline that was engineered can be engineered differently. That is the first thing the deficit model obscures.

Every Institution Encodes a Prototype

Here is the concept the rest of the essay turns on. Every institution is built around an implicit model of the standard human being it is meant to serve. Call it the prototype: the default user the system was designed around, the person whose capacities and tendencies the architecture quietly assumes. The prototype is rarely written down; it is embedded in the design itself, in the form of what the system takes for granted. And it is specific in every institution.

Schools encode a prototype learner: one who can sit still, attend to spoken instruction, tolerate a particular sensory load, transition on command, and intuit the unwritten rules of the room. A child who matches is “well-behaved”; a child who does not is recorded as a behavior problem, and the record turns punitive fast. In the 2020-2021 U.S. federal civil-rights data, students with disabilities accounted for roughly 81 percent of those physically restrained and about 75 percent of those secluded, despite being around an eighth of enrollment (U.S. Department of Education, 2024). The autistic child who leaves a room because of sensory overload is recorded as defiant and punished, even though leaving is a direct manifestation of the disability the school is obligated to accommodate (EdTrust, 2025). The design produces the behavior; the behavior is attributed to the child; the attribution justifies the design.

Workplaces encode a prototype employee: one who interviews well, networks fluently, reads tone, and performs the culturally specific signals of confidence and fit. Autistic adults, fully capable of the work, are screened out at rates that show up in national statistics: the UK’s Office for National Statistics found only about 29 percent of autistic adults in work, among the lowest of any disability group, and the 2024 Buckland Review confirmed only about three in ten working-age autistic people are employed despite the majority wanting to work (ONS, 2022; Buckland, 2024).

Healthcare encodes a prototype patient: one whose distress is legible to standardized instruments and whose improvement is defined as movement toward the norm. When instruments measure symptom reduction rather than well-being, a patient who suppresses visible autistic traits looks “improved” even as their mental health declines under the strain (Cassidy et al., 2018; Raymaker et al., 2020).

Courts encode a prototype witness: one whose credibility is read off demeanor, the steady eye contact, the appropriate emotional expression, the fluent speech. A mind that produces a different demeanor is read not as different but as dishonest. In one controlled study, mock jurors shown an autistic defendant with no explanation perceived him as deceitful and were more likely to convict; shown the same defendant with his autism explained, the same jurors rated him as more honest and convicted less often (Maras, Marshall & Sands, 2019). Same person, same behavior, opposite verdict. The only variable was whether the system could recognize the mismatch for what it was.

And beneath all of these, society encodes a prototype human: the assumed default against which everyone is implicitly compared, and which most people match closely enough never to notice. Which raises the question the rest of the essay presses: what happens when the prototype becomes invisible?

The Invisible Prototype

The answer is that an invisible prototype stops looking like a design choice and starts looking like reality itself.

This is the quiet engine of the whole system: the most powerful assumptions are the ones that appear neutral. A prototype that is named can be debated. A prototype that has vanished into “just how things are” cannot, because no one experiences it as a choice. The staircase does not feel like a position on who belongs in the building; it feels like a floor going up. The interview does not feel like a theory of competence; it feels like meeting someone. The diagnostic manual does not feel like a definition of the standard mind; it feels like a description of medicine. The assumption becomes invisible precisely to the degree that it is widely shared.

And here is the asymmetry that makes the prototype durable. The people who match it cannot see it. For them there is no friction, no gap between themselves and the design; the environment fits like air. Only the people excluded by the prototype can perceive it, because they are the ones who collide with it. The misfit is not a flaw in their perception; it is a form of evidence available only to them. The wheelchair user knows where every missing ramp is. The autistic person knows exactly which unwritten rule just cost them the room. They are, in a precise sense, the system’s only honest instruments: the sensors that register the assumptions everyone else is standing inside of.

The communication research makes this concrete. The diagnostic criteria frame autistic social difficulty as a one-way deficit, but Damian Milton’s “double empathy problem” reframed it as a reciprocal mismatch: non-autistic people are as poor at reading autistic people as the reverse (Milton, 2012). When Crompton and colleagues tested it, information passed accurately down chains of autistic people and down chains of non-autistic people, and degraded only in the mixed chains (Crompton et al., 2020). The breakdown was never located in one group; it lived in the gap between them. But because the non-autistic style is the prototype, the gap gets charged entirely to the side that does not match, and the side that does match never has to notice there was a gap at all.

This is what it means to say that normal is not neutral. Neutrality is the costume a dominant assumption wears once it has become invisible. The language stays mild, full of words like standards, competencies, fit, and professionalism, while the outcomes sort along the contour of a prototype no one will admit to having drawn.

Social Legibility and Institutional Legibility

Once the prototype is visible, a great deal of what institutions describe as measuring ability turns out to be measuring something else. Institutions believe they are assessing competence. Often they are assessing compatibility with their own design assumptions, and reporting the result as if it were competence. The two come apart wherever a proxy stands in for the thing itself. The interview is a proxy: it measures the capacity to signal competence in a culturally specific style and treats the signal as the substance. The standardized classroom measures compliance with a particular mode of attention and records it as learning. The credibility heuristic measures demeanor and reports it as honesty. In each case the system mistakes fit with the prototype for the underlying capacity. Institutions, in short, routinely mistake conformity to the prototype for competence, and because they are optimized for sameness, the mistake is invisible to them. The system measures fit, not ability.

It helps to separate two layers at which this reading happens, because they fail independently. Social legibility is interpersonal: whether one person can readily read and understand another, as when an interviewer reads a candidate or a juror reads a witness’s demeanor. Institutional legibility is systemic: the degree to which a person can be recognized, interpreted, and validated by an institutional apparatus. It asks whether the diagnostic manual has a category that fits them, whether the eligibility form has a box they can truthfully check, whether the credentialing system can register what they actually know, whether the hiring algorithm can score them as anything other than a deviation. Social legibility asks whether you can be read by a person. Institutional legibility asks whether you can be read by a system. Both are measured against the same prototype, and a person can pass one while failing the other: legible to a sympathetic manager who understands them perfectly, yet illegible to the form, the test, or the model that actually decides who qualifies. It is institutional legibility, more than its interpersonal cousin, that determines who is employable, who is believed, who is eligible, and who is recognized, because the system, not the sympathetic individual, holds the authority to grant or withhold.

The newest enforcement of legibility is automated, and it shows the prototype hardening into infrastructure. Many employers now screen candidates with AI video-interview tools that score facial expression, eye contact, and vocal patterns. In 2022 the U.S. Equal Employment Opportunity Commission and Department of Justice jointly warned that such tools may unlawfully screen out qualified people with disabilities; the Department of Justice’s guidance gives the explicit example that facial and voice analysis may screen out qualified applicants with disabilities such as autism (U.S. Department of Justice, 2022; EEOC, 2022). What was once an interviewer’s implicit bias is now a measurable variable in a model. The proxy has not been retired; it has been automated, which is to say institutional legibility now operates at machine scale.

Culture Fit and Professionalism

The purest everyday instance of legibility mistaken for ability has a familiar name: culture fit. Organizations believe culture-fit hiring selects for talent, capability, and performance. The evidence says it frequently selects for something else. In a landmark study of elite professional-service firms, the sociologist Lauren Rivera found that hiring functioned as cultural matching: evaluators sought candidates who were not only competent but culturally similar to themselves in leisure pursuits, background, and self-presentation, and more than half ranked cultural fit as the single most important criterion at the interview stage, a concern that often outweighed productivity itself (Rivera, 2012). Hiring, she concluded, resembled the selection of friends more than the selection of workers.

Strip the euphemism and culture fit is institutionalized social preference: a hiring system optimized for familiarity, comfort, and predictability, presented as a judgment of merit. “Professionalism” works the same way. It names a specific register of dress, speech, affect, and bodily comportment, from eye contact to the right kind of small talk to the right emotional display, and treats fluency in that register as evidence of capability. None of these signals is the work. They are the signals of resemblance to the prototype, and a candidate who has the capability but not the resemblance is read as unprofessional, a poor fit, somehow not quite right: a verdict that feels like a perception of the person but is in fact a measurement of distance from a cultural default no one will name.

The Translation Tax

For those who do clear these filters, the demand does not end. It goes underground, and it has a cost. To remain legible inside institutions built around a different prototype, many autistic adults perform continuous, deliberate translation: rehearsing scripts, masking discomfort, suppressing natural movement, decoding unwritten rules, converting their own thoughts into the institutionally acceptable format before speaking. This is the translation tax: the hidden, unpaid, uncounted labor of manufacturing a resemblance one does not naturally possess.

It means that many autistic people are effectively working two jobs at once: the actual job, and the job of continuously translating themselves into the institution’s language so they can keep doing the first one. The tax is invisible on every ledger. It does not appear in performance reviews, which record only the output, not the cost of producing it. But it appears, eventually, in the body. Camouflaging is robustly associated with anxiety, depression, exhaustion, and a documented phenomenon of autistic burnout, and was identified as a specific risk marker for suicidality, against a backdrop in which up to roughly 79 percent of autistic adults meet criteria for a co-occurring psychiatric condition and up to two-thirds report lifetime suicidal thoughts (Cassidy et al., 2018; Raymaker et al., 2020). The prototype does not only exclude the people who cannot perform it. It quietly bills the people who can.

The Normalization Pipeline

The translation tax is not something individuals improvise alone. It is taught, across a sequence of institutions, as an explicit program, what might be called the normalization pipeline. Its origin point is unusually well documented. The foundational study of intensive early behavioral intervention for autism, published by Ivar Lovaas in 1987, defined its own success as making autistic children “indistinguishable from their peers” (Lovaas, 1987). Indistinguishability, not flourishing, not communication, not well-being, was the stated goal. The objective, from the start, was to reduce the visible distance from the prototype.

That logic does not stop in early childhood; it runs the length of a life, handed from institution to institution. Early behavioral intervention teaches the child to suppress visible difference. School discipline teaches conformity to the prototype learner. Social-skills training teaches the performance of neurotypical interaction. Interview coaching teaches the manufacture of social legibility on demand. Performance management and “professionalism” expectations enforce it through adulthood. Each institution operates independently, with its own vocabulary and rationale, and each delivers a version of the same instruction: become more legible to the dominant system. Normalization is usually imagined as something that happens to children and then ends. In fact it is a lifelong institutional process, a relay in which the baton is the same demand passed from hand to hand, and the translation tax is simply the bill for complying with it after the formal training stops.

The Friction Framework

Follow one person through these institutions and a pattern resolves that no single institution can see. The same kinds of people encounter the same kinds of friction across systems that have no connection to each other. The school, the hiring process, the clinic, and the courtroom were designed by different people for different purposes in different centuries. They share no committee, no blueprint, no coordination. And yet they fail the same person in structurally similar ways, each reading a mismatch with its own prototype as a defect in the individual.

That repetition is the most important piece of evidence the system produces, and it is invisible from inside any one institution. Each sees an isolated case: one difficult student, one poor interview, one non-compliant patient, one unconvincing witness. Only when you aggregate across institutions does the pattern resolve. The same person being read as deficient by four unrelated systems is not four independent data points about that person. It is one data point about the prototype that all four systems happen to share.

So the legitimacy question sharpens: at what point does repeated friction stop being evidence of personal weakness and start becoming evidence of design assumptions? There must be such a point. A single failed interview is plausibly about the candidate. A lifetime of friction across every standardized institution a person passes through, all pointing the same direction, is not plausibly about the candidate. At some threshold the explanation must migrate from the individual to the architecture, because no theory of individual deficiency can explain why the deficiency appears identically in systems that never spoke to one another.

This is where the personal and the structural blur, and why the confusion is so costly. What feels personal is often structural. The friction arrives one institution at a time, in the first person, addressed to you by name: your suspension, your rejection, your diagnosis, your failure to convince. It does not announce itself as systemic. So the person on the receiving end does the natural thing and concludes that the common factor is themselves, internalizing as private failure what is in fact the signature of a shared design. The prototype’s final trick is that it makes its own subjects into expert witnesses against themselves.

Institutional Legitimacy and the Legitimacy Loop

This is the point at which the argument stops being only about autism. The deepest function of an institution is not to provide a service. It is to confer legitimacy, to decide who counts as a competent worker, a credible witness, a reliable patient, a successful person, a full participant. This is the concern of what I have called Institutional Legitimacy™: institutions hold the authority to define the categories of social worth and then to sort people into them, exercising that authority through prototypes most people never see, and recording the costs of their own narrow design as the failings of the people it excludes.

Set out plainly, the mechanism is a chain. An institution defines a prototype. Legibility, social and institutional, is measured against that prototype. Legitimacy is granted to those who resemble it closely enough to be read as competent, credible, or qualified. And access follows legitimacy: to the job, the diagnosis, the benefit of the doubt, the place at the table. Everyone the prototype did not anticipate pays a translation tax to manufacture the resemblance. Prototype, legibility, legitimacy, access. Each link looks like a neutral judgment of the individual, and the chain as a whole looks like meritocracy. It is, in large part, a measurement of proximity to a standard no one named.

What makes the system durable is that it reproduces itself, and it does so without anyone intending it to. Call it the legitimacy loop. Institutions define normal. Systems reward those who match it. The rewarded rise into the positions where decisions are made. From there, they reinforce the definition of normal that selected them, not from malice but from the ordinary human tendency to recognize competence in people who resemble oneself. Divergence, measured against a prototype now staffed and defended by the people it favored, appears more deficient than ever. And the loop closes and runs again. This is why the same pattern appears across education, healthcare, employment, law, media, and artificial intelligence without any coordination between them. There is no conspiracy. There does not need to be. The structure reproduces itself, because the people it elevates are precisely the people least able to see it.

The Economic Consequences of Exclusion

It is tempting to treat all of this as an ethical matter, and it is one. But the prototype also has a price, and the price is paid by everyone, which moves the argument out of ethics and into public policy. A system optimized for sameness does not merely treat outliers unfairly. It forecloses the capability it cannot read, and forgone capability is an economic loss.

The macro estimates are necessarily rough, but they point one direction. An International Labour Organization working paper estimated that excluding people with disabilities from the labor market costs economies in the range of 3 to 7 percent of GDP, based on a study of ten low- and middle-income countries (Buckup, 2009). Its author drew the conclusion this essay has been building toward: the lost productivity reflects disabling environments rather than disabled people, because the cost arises from how work is designed, not from the workers. At the level of the firm, an analysis by Accenture with Disability:IN and the American Association of People with Disabilities reported that companies leading on disability inclusion outperformed peers financially over the period studied, posting higher revenue, higher net income, and higher economic profit margins (Accenture, Disability:IN & AAPD, 2018). And the redesigned hiring programs discussed below were documented in the Harvard Business Review not as charity but under the banner Neurodiversity as a Competitive Advantage, on the argument that the belief that scalable work requires conformity to standardized processes is itself what screens out capable people (Austin & Pisano, 2017).

The figures vary and should be read as orders of magnitude rather than precise accounting. But the structural claim does not depend on the exact number. A society that builds its institutions around a narrow cognitive prototype, screens out everyone who does not match, and then interprets the resulting exclusion as individual deficiency is not running an efficient meritocracy that happens to have some unfortunate casualties. It is systematically discarding capability it has disqualified itself from recognizing, and then paying for the discard twice: once in foregone production, and again in the public cost of supporting the people it shut out.

Mismatch Is Data, Not Deficiency

These observations converge on an alternative to the deficit model. Where the conventional account treats the autistic person as a fixed bundle of deficits and the institution as a neutral backdrop, this account reverses figure and ground. The institution is a designed artifact encoding a prototype; the autistic person is a real and valid mind; and the deficits the institution records are, in large part, a measurement of the mismatch between the two, read as a property of the person because the institution cannot see its own design. Repeated friction is not a verdict on the individual. It is information about the fit. Mismatch is data, not deficiency.

The strongest test of the claim is the natural experiment: hold the people constant and change the environment. If the difficulty is purely internal, redesign should do little; if it is produced at the interface, redesign should move the outcomes. Across domains, it moves them. When SAP replaced conventional interviews with multi-week, task-based assessments in its Autism at Work program, capability the interview had been screening out became visible, and the program reports retention around 90 percent across hundreds of employees in more than a dozen countries (SAP, 2024). Match the neurotype and the communication “deficit” substantially disappears (Crompton et al., 2020). The Buckland Review reached the same conclusion in the language of policy, recommending a shift from individual accommodation toward inclusion “by design”: changing recruitment and workplace practices universally so the benefit does not depend on each autistic person first disclosing and requesting help (Buckland, 2024).

This is the principle of universal design, developed in the built environment by the architect Ronald Mace: design for the full range of human variation from the start rather than building for a prototype and retrofitting exceptions. Its signature example is the curb cut, built for wheelchair users and used in the end by nearly everyone. The accommodation, generalized, became a better design for the whole population. The same holds for cognition. Naming these mechanisms is not a denial that autistic people face real difficulty, nor a fantasy that redesign is costless or able to erase every support need. It is an insistence on an honest accounting of where difficulty originates, because the accounting determines the response. If the problem is wholly internal, the only available interventions are to fix, reduce, or manage the person, which is what we have overwhelmingly funded: UK autism research for 2007 to 2011 directed 56 percent of expenditure to biology and cognition and just 1 percent to the societal conditions autistic people actually live inside, almost the reverse of the priorities the community itself reported (Pellicano, Dinsmore & Charman, 2014). If a substantial share of the difficulty is a design mismatch, a whole category of interventions we have barely tried becomes available: changing the design.

Autism as a Systems Diagnostic

So we can name, finally, what autism is doing in this argument. It is not the subject. It is the instrument.

Autism functions here as a systems diagnostic: a way of revealing assumptions hidden inside institutions, the way a stress test reveals the weaknesses in a bridge. The stress test does not break the bridge; it exposes where the bridge was already weak. Autistic experience does the same to institutions: by failing to match the prototype, it makes the prototype visible, and by colliding with the design, it locates exactly where the design assumed a narrower humanity than it admitted to. The same logic that mistakes autistic legibility for autistic incompetence operates wherever a system encodes a narrow default and treats distance from it as deficiency: in who gets read as articulate, who gets read as professional, who gets believed, and who gets counted as having succeeded. Autism is a sharp case because the mismatch is sharp. It is not a special case.

This reframes the entire inquiry. We have spent a century measuring autistic people against a standard we never named, recording their distance from it as impairment, and assuming the distance lived inside them. The honest reckoning is that we cannot say it does. Autism, looked at clearly, turns out not to be primarily a deficit problem at all. It is a design problem: a mismatch we built, mislabeled, and then charged to the people it excludes. So the question we began with should be turned all the way around. The question is not whether autistic people can adapt to the systems we have built. The question is what autistic experience reveals about those systems: about the prototype they were built around, the legibility they demand, the legitimacy they distribute, and the variation they were never designed to hold.

When the same experiences are repeatedly dismissed across different people and contexts, that is not randomness. That is structure.

© 2026 Eva Redford

References

Accenture, in partnership with Disability:IN and the American Association of People with Disabilities. (2018). Getting to Equal: The Disability Inclusion Advantage.

Asperger, H. (1944). Die “Autistischen Psychopathen” im Kindesalter. Archiv für Psychiatrie und Nervenkrankheiten, 117, 76-136.

Austin, R. D., & Pisano, G. P. (2017). Neurodiversity as a competitive advantage. Harvard Business Review, 95(3), 96-103.

Buckland, R. (2024). The Buckland Review of Autism Employment: Report and Recommendations. Department for Work and Pensions, HM Government.

Buckup, S. (2009). The price of exclusion: The economic consequences of excluding people with disabilities from the world of work (Employment Working Paper No. 43). International Labour Office, Geneva.

Cassidy, S., Bradley, L., Shaw, R., & Baron-Cohen, S. (2018). Risk markers for suicidality in autistic adults. Molecular Autism, 9, 42. https://doi.org/10.1186/s13229-018-0226-4

Centers for Disease Control and Prevention. (2025). Data and statistics on autism spectrum disorder (Autism and Developmental Disabilities Monitoring Network, surveillance year 2022). U.S. Department of Health and Human Services.

Crompton, C. J., Ropar, D., Evans-Williams, C. V., Flynn, E. G., & Fletcher-Watson, S. (2020). Autistic peer-to-peer information transfer is highly effective. Autism, 24(7), 1704-1712. https://doi.org/10.1177/1362361320919286

Davis, L. J. (1995). Enforcing Normalcy: Disability, Deafness, and the Body. Verso.

EdTrust. (2025). Unfair punishments: How school discipline disproportionately targets students with disabilities.

Equal Employment Opportunity Commission. (2022). The Americans with Disabilities Act and the use of software, algorithms, and artificial intelligence to assess job applicants and employees (Technical assistance document, May 12, 2022).

Kanner, L. (1943). Autistic disturbances of affective contact. Nervous Child, 2, 217-250.

Lotter, V. (1966). Epidemiology of autistic conditions in young children: I. Prevalence. Social Psychiatry, 1, 124-137.

Lovaas, O. I. (1987). Behavioral treatment and normal educational and intellectual functioning in young autistic children. Journal of Consulting and Clinical Psychology, 55(1), 3-9.

Maras, K., Marshall, I., & Sands, C. (2019). Mock juror perceptions of credibility and culpability in an autistic defendant. Journal of Autism and Developmental Disorders, 49(3), 996-1010. https://doi.org/10.1007/s10803-018-3803-7

Milton, D. E. M. (2012). On the ontological status of autism: The “double empathy problem.” Disability & Society, 27(6), 883-887. https://doi.org/10.1080/09687599.2012.710008

Office for National Statistics. (2022). Outcomes for disabled people in the UK: 2021. ONS.

Pellicano, E., Dinsmore, A., & Charman, T. (2014). What should autism research focus upon? Community views and priorities from the United Kingdom. Autism, 18(7), 756-770. https://doi.org/10.1177/1362361314529627

Quetelet, A. (1835). Sur l’homme et le développement de ses facultés, ou Essai de physique sociale. Bachelier.

Raymaker, D. M., Teo, A. R., Steckler, N. A., Lentz, B., Scharer, M., Delos Santos, A., Kapp, S. K., Hunter, M., Joyce, A., & Nicolaidis, C. (2020). “Having all of your internal resources exhausted beyond measure and being left with no clean-up crew”: Defining autistic burnout. Autism in Adulthood, 2(2), 132-143. https://doi.org/10.1089/aut.2019.0079

Rivera, L. A. (2012). Hiring as cultural matching: The case of elite professional service firms. American Sociological Review, 77(6), 999-1022. https://doi.org/10.1177/0003122412463213

SAP. (2024). The win-win potential of hiring neurodiverse workers and Autism at Work program reporting. SAP SE.

U.S. Department of Education, Office for Civil Rights. (2024). Civil Rights Data Collection: 2020-21 school year(restraint, seclusion, and discipline findings).

U.S. Department of Justice, Civil Rights Division. (2022). Algorithms, artificial intelligence, and disability discrimination in hiring (Guidance, May 12, 2022).

Wing, L. (1981). Asperger’s syndrome: A clinical account. Psychological Medicine, 11(1), 115-129.

Wing, L., & Gould, J. (1979). Severe impairments of social interaction and associated abnormalities in children: Epidemiology and classification. Journal of Autism and Developmental Disorders, 9(1), 11-29.

Editor’s note: This essay is part of the Institutional Legitimacy series, which examines the gap between what institutions claim to do and what their design actually does. This piece looks at what autistic experience reveals about the hidden assumptions embedded in our major institutions — and why that signal keeps getting misread.

Every April, we talk about autism.

We hold awareness events. We share educational posts. We encourage people to recognize autistic traits.

And most people do recognize them now.

So here’s the uncomfortable question no one’s asking:

What if awareness of autism was never really the problem?

What if the real problem is that we’re not aware of the environments autistic people are forced to survive inside?

Most of our major institutions — schools, offices, hiring systems, social norms — were designed before anyone seriously considered neurological diversity.

They were built around silent assumptions: about how people think, how they communicate, how they manage attention.

Over time, those assumptions became invisible. They became what we call normal.

But normal isn’t neutral. Normal is a design outcome.

Here’s what happens when autistic people enter these environments: the hidden assumptions suddenly become visible.

Noise levels that never stop. Social rules that must be inferred, not explained. Rapid task-switching. Constant performance under observation.

For some people, these conditions feel unremarkable. For others, they create relentless sensory and cognitive friction.

And when that friction appears — when someone struggles — society almost always reaches the same conclusion: the individual must be the problem.

There is another possibility we rarely consider.

Autistic people may be stress-testing our systems. And revealing where the design breaks.

In engineering, you stress-test a system not to destroy it — but to find where its assumptions no longer hold.

That is exactly what autistic experience is doing to our institutions.

Not exposing a flaw in the person. Exposing a flaw in the design.

Look at what schools actually ask children to do all day.

Sit still for hours. Process language at speed. Shift attention on demand. Tolerate constant noise. Navigate unwritten social rules. Recover instantly from correction.

None of this is on the syllabus.

But children who do it effortlessly are called good students. Children who can’t are called behavior problems.

For many autistic students, the hardest part of school isn’t learning the material. It’s surviving the invisible curriculum wrapped around it.

By adulthood, the environment changes — but the assumptions don’t.

Workplaces reward people who can multitask constantly, read between the lines in every meeting, tolerate chaotic sensory environments, and perform social ease as a professional skill.

These traits get labeled professionalism.

But professionalism — like normal — is shaped by the system that defined it. Not by the full range of human capability.

The awareness conversation is ready to evolve.

The first phase asked society to recognize that autistic people exist. That mattered.

But the next question is harder — and more important: what do autistic experiences reveal about the systems themselves?

When the same environments consistently produce burnout, distress, and exclusion for large groups of people, that is not just an individual problem. That is a design problem.

Autistic people often experience environments more intensely, more immediately, more visibly. That sensitivity is frequently framed as a deficit.

But it can also function as a signal — one that exposes sensory overload designed into classrooms, social ambiguity baked into professional culture, hiring systems built for conformity over capability, institutions optimized for uniformity over human variation.

Autistic experience works like a diagnostic tool for systems. It reveals where the environment is poorly calibrated for the range of human minds that actually exist.

So what does autism awareness look like in 2026?

Maybe it’s no longer just about recognizing autism.

Maybe it’s about recognizing how institutions quietly define who counts as normal — and who gets excluded by that definition.

The people who struggle most inside a system aren’t always the ones who need to change.

Sometimes they’re the ones showing us where the system itself needs to.

When the same experiences are repeatedly dismissed across different people and contexts, that is not randomness. That is structure.