7 min read

Two Sides of the Same Coin

neurodivergentaccessibilityphilosophy

Two students open the same study app. One has spent years being failed by software that assumed a brain like everybody else's. The other is three weeks out from the single exam that decides the next decade of their life. On paper they want opposite things. In practice they need almost exactly the same tool, and this post is about why.

The two who look nothing alike

The first is the neurodivergent learner. ADHD, dyslexia, dyscalculia, autism. Written down, their needs read as a list of accommodations: less on the screen, nothing moving suddenly, a way in that doesn't depend on parsing a wall of text, a session that survives a day where their brain flatly refused to start.

The second is the high-stakes overachiever. The medical student, the trainee lawyer, the candidate whose entire future is riding on one result. Written down, their needs read as the opposite of that. Not less, more. More depth, more rigour, more hours, more certainty that every single fact is right.

So one appears to need the tool to carry them and the other appears to need it to get out of the way, and it's tempting to conclude from that you'd have to build two different products. That conclusion is the mistake.

They're failed by the same middle

The reasons each one is underserved rhyme, which is the first clue.

The neurodivergent learner got failed first, by mainstream tools built around a student who reads quickly, starts easily and doesn't get derailed by a notification chime. When a tool assumes a brain you haven't got, you go and look for one that works, and that search is a big part of why neurodivergent learners are among the earliest and most committed adopters of AI study tools. Not novelty. Necessity.

The overachiever gets failed from the other end. Their defining feature is what a mistake costs them: a misremembered mechanism carried into an exam that gates a career isn't a dropped mark, it's a dropped year. And yet the tools they end up reaching for are optimised around a median user cramming for a quiz, not around somebody whose margin for error is zero.

The same three failures land on both

Look at how the current crop of tools fails and it's the same three faults over and over.

Overstimulation. Streaks, points, confetti, a chime on every correct answer, game footage looping behind the text you're supposed to be reading. Mayer's (2009) coherence principle is blunt about this: adding interesting-but-irrelevant material to a lesson reliably reduces how much gets learned. And consistent with LaBar and Cabeza (2006), the arousal that genuinely helps memory is moderate and intrinsic to the task, not a burst of external celebration bolted onto the end. For an autistic or ADHD student that motion and sound isn't merely unhelpful, it breaks focus and overloads a working-memory system that was already stretched. The reward layer also fails on its own terms, because Deci, Koestner and Ryan (1999) found that attaching extrinsic rewards to something a person already wanted to do erodes the internal motive, and Hanus and Fox (2015) watched gamification drag down both intrinsic motivation and final-exam performance across a full course. And streaks carry a particular cruelty for anyone whose capacity swings day to day, since Barkley (1997) describes the ADHD missed day as neurological rather than motivational, which means the streak is punishing a symptom.

Noise and decision-load. A dashboard with nine things to click is nine decisions standing between you and any learning. Sweller's (1988) cognitive-load theory rests on working memory holding only a handful of items at a time, and every non-essential control, badge and menu spends part of that budget on the interface rather than the material. For a learner already carrying an executive-function load, deciding what to do next is frequently the thing that ends a session before it begins. For the overachiever it's a friction tax on time they haven't got. A Thousand Small Decisions goes through both properly.

Shallow, do-it-for-you tooling. Ask the newest apps a question and you get a finished answer back, but the answer was never where the learning lived. Slamecka and Graf (1978) showed that information a learner generates themselves is remembered better than the same information read, and Roediger and Karpicke (2006) showed that retrieving a fact does more for durable memory than rereading it does. So a tool that does the thinking on your behalf has removed the precise act that was building the memory. That robs the overachiever of the deep work their exam is going to demand, and it robs the neurodivergent learner too, who needed a scaffold to help them do the work, not a machine to do it in their place.

Invert those and one design covers both

Now take each failure and turn it over.

No gamification, because the calm that lets an autistic student stay in a session is the same calm that lets a medical student concentrate for three hours straight. Predictable layouts, because a screen that never rearranges itself saves an ADHD learner a re-orientation tax and saves everyone else a distraction. One clear next step, because taking the decision away frees executive function for the first student and protects deep focus for the second. Voice input everywhere, because a dyslexic student who thinks faster than they type and a time-pressed finalist who wants to reason out loud are helped by the identical feature. Sessions that resume and don't punish a missed day, because variable capacity is a fact of ADHD and also just a fact of having a life. And accuracy you can actually rely on, because the student who can't afford a wrong fact and the student who wouldn't easily catch one both need a tool that doesn't quietly make things up.

None of that overlap is a coincidence. It's the oldest principle in accessible design, the curb cut lowered for wheelchairs that everybody dragging a suitcase now uses, and CAST's (2018) Universal Design for Learning framework formalises it: design for the margins and the centre benefits anyway.

The tie-break

When a feature would help the comfortable student but harm the neurodivergent one, the neurodivergent student wins, and the feature is removed rather than buried in a settings menu. Build to that rule and you don't get a narrower tool, you get a calmer and more honest one for everybody.

That rule gets argued in full in Calm by Design, and it's the thing stopping these two audiences from pulling the design in half. They never do, because they were never pulling in opposite directions. They're asking, in two quite different accents, for the same three things: accuracy they can trust, calm that doesn't overwhelm them, and enough structure that deciding what to do next stops being their problem.

Build genuinely for the student the average tool forgets, and you've built for the student who can't afford for it to fail.

Sources

Read more