# ADHD Cognition → Design Rules (research digest) Why the Cognitive Contract in SKILL.md §1 is shaped the way it is. Each finding maps to a build rule. This digest targets the interest-based learner who runs multiple threads simultaneously, loves understanding things, and needs the understanding to *stick*. --- ## 1. The interest-based nervous system (Dodson) ADHD motivation is not driven by importance or consequences; it is driven by dopamine-releasing task properties. The popular mnemonic is **PINCH**: - **P**lay / Passion — the task is intrinsically enjoyable - **I**nterest — inherently fascinating framing - **N**ovelty — new format, new angle, new tool - **C**hallenge / Competition — just-hard-enough stretch; prediction games - **H**urry (urgency) — finite, bounded, deadline-shaped **Build rules:** - Frame the topic through what the reader already finds fascinating (Topic Read step 2 — anchor to *their* world). - Novelty is a *renewable* resource: each chunk shift is a format shift (animation → diagram → interactive → screenshot). Two identical formats adjacent = attention leak. - Challenge via prediction checkpoints: "what happens if X doubles?" → tap a guess → immediate animated reveal. Just-hard-enough; never scored; wrong guesses get "most people guess that — here's the twist" not a red X. - Urgency via *boundedness*, not pressure: "chunk 3 of 7", visible finish line, each chunk ≤90s. Open-ended walls of content read as un-startable. ## 2. Front-loaded attention Engagement in ADHD learners peaks at the very start of any session and decays; the hook must come first, not mid-lesson. Classic teaching saves the payoff; this is exactly backwards for this audience. **Build rules:** - First viewport carries the most surprising fact, a live phenomenon, or the "why should I care" stake. Definitions, history, and prerequisites come *after* curiosity is committed. - Cold-open pattern: show the *effect* first (the animation running), explain the *cause* second. Mystery → mechanism, never mechanism → so-what. ## 3. Working memory constraints ADHD working memory is smaller and more easily disrupted. Every item the page asks the reader to "hold in mind" is a tax. Losing one's place after a distraction (which WILL happen) is the standard failure. **Build rules:** - Externalize state: persistent progress rail = mini-map of the ideas, current position highlighted. Re-entry after distraction costs zero reconstruction. - Chunk to ~3 items: never ask the reader to track more than 3 new named things per chunk. If the mechanism has 6 parts, pre-train the parts (names + one-line roles, visually) in one chunk, then the interactions in the next. - Keep referents visible: when text says "the resolver", the resolver is highlighted in the diagram *at that moment* (scroll-synced or hover-synced highlighting). Never make the reader scroll back. - One idea per screen. A screen that requires holding the previous screen in memory is two ideas. ## 4. Cognitive Load Theory (Sweller) + Mayer's multimedia principles Working memory budget = intrinsic load (the topic's real difficulty) + extraneous load (everything else). You cannot reduce intrinsic load, only manage it (segmenting, pre-training); extraneous load you cut ruthlessly. The empirically strongest levers (Mayer): - **Coherence** — excluding extraneous material improves deep learning; the strongest and most replicated effect (median effect size ~0.86 across 23/23 tests). Decorative images, tangents, and mood-content actively harm learning. - **Signaling** — visually cue what to look at (highlight, arrow, spotlight-dim-the-rest) exactly when it matters. - **Segmenting** — learner-paced segments beat continuous presentation. The reader controls advance. - **Spatial contiguity** — labels ON the diagram, not in a caption or paragraph. Distance between word and picture is load. - **Temporal contiguity** — the words about a motion appear *while* the motion plays. - **Multimedia / dual coding** — words + pictures beat words alone; verbal and visual channels are separate capacity. Encoding an idea in both doubles retrieval paths. - **Pre-training** — name the components before explaining the process. - **Personalization** — conversational "you" beats formal register. **Build rules:** these are directly §1 rules 2, 3, 5, 8. When in doubt, coherence wins: cut it. ## 5. Immediate feedback loops Delayed reward has weak pull on ADHD motivation; feedback must be immediate, frequent, and specific. This is why passive video courses fail this audience and why interaction cadence matters. **Build rules:** - Every interactive element responds in <100ms with a visible consequence. - Checkpoint interactions every 2–3 chunks; completing a chunk gets a small, honest completion signal (progress rail ticks, subtle check) — not confetti, not points, no streaks. Gamification theater reads as manipulation; *progress made visible* reads as respect. ## 6. Hyperfocus and the rabbit-hole problem The same brain that can't start a boring page will tunnel for an hour into a fascinating tangent. Tangents mid-mechanism derail the main thread; but killing them wastes the audience's superpower. **Build rules:** - Park every tangent in a collapsible **Rabbit Holes** drawer at the end (or a small margin marker linking to it). The reader finishes the spine of the explanation, *then* tunnels. - Design the spine to be finishable in one sitting (5–9 chunks); depth lives in the drawer. ## 7. Multi-threaded learners This audience often learns several things simultaneously and connects across domains. Cross-domain analogy is not a crutch — it's the native encoding format. **Build rules:** - Analogies drawn from the reader's *other* active interests stick hardest (use conversation context/memory). - The recap card's job is cross-linking: "this is the same shape as X you already know" is a legitimate, powerful anchor. - Vivid/absurd anchors beat sensible ones for recall — one deliberately weird image per key idea in the recap ("DNS is a chain of increasingly specific gossips"). Weird sticks; bland evaporates. ## 8. Reading mechanics - Continuous prose > ~60 words per block starts skipping. Short paragraphs, high-signal bolding (1–2 phrases per block max — over-bolding destroys signaling). - Sans-serif or high-legibility type at generous size (≥17px body), line length 50–70ch, real line-height (1.6+). - Motion as *involuntary attention capture* is a double-edged sword: purposeful motion directs attention (good); ambient looping motion captures it continuously (catastrophic for this audience). Nothing loops while the reader is meant to be reading. - Session length respect: the page must be pausable/resumable at any chunk with zero penalty. ## 9. Emotional register (rejection sensitivity & anxiety) Many ADHD adults carry rejection sensitivity and shame around "should already know this". Anxiety-adjacent topics amplify this. **Build rules:** - Never "simply" / "just" / "obviously" / "as everyone knows". The tone is a sharp friend explaining, not a professor grading. - Prediction checkpoints are framed as games, never tests. No scores anywhere. - Anxiety-adjacent topics (medical, debt, legal): calm palette, slower motion, explicit uncertainty ranges, no urgency theater. --- ## Honest caveat Direct empirical research on *which specific design elements* help ADHD comprehension is thin; most of the above extrapolates from working-memory research, cognitive load theory, multimedia learning research (robust), and clinical/educational practice consensus (moderate). Where practice consensus and Mayer conflict, Mayer wins. These principles also measurably help neurotypical readers under fatigue/distraction — designing for the ADHD reader is designing for everyone's worst afternoon.