A new paper just published in AI & Society puts the AIR framework research transparency model under a microscope—and the results are a mixed bag. The framework is a genuine leap forward for honest AI disclosure in research. But it also has five significant gaps, and one of them could quietly punish disabled and neurodivergent researchers for using the very tools that make their work possible.

You are not failing if your autistic or ADHD child doesn’t look “normal.” A child who rocks, flaps, scripts, or info‑dumps and actually feels safe is healthier than a child who sits still, smiles on cue, and is quietly disintegrating inside.

For decades, most “autism research” has quietly treated autistic people as problems to be managed, not as people whose lives should get better. The primary outcome has been behavior control for neurotypical comfort, not autonomy or quality of life for autistic communities themselves.

Most “autism tech” isn’t actually built to make autistic lives safer or less painful. It’s built to watch—tracking gaze, movement, and “engagement” so adults can see dashboards and funders can see metrics, while the autistic person absorbs the risk and rarely gets more control.

Many autistic adults now describe their early “intensive therapy” as ABA trauma: years of compliance training and stimming suppression that left them with PTSD‑like symptoms and a deep mistrust of their own bodies. This piece explores how that happens, and what care would look like if we put autistic people’s safety—not adult comfort—at the center.

We keep congratulating autistic people for “masking well” while the research quietly piles up: camouflaging autistic traits is tied to depression, anxiety, burnout, and suicidality. This isn’t a quirky social skill. For many people, it’s a slow, socially rewarded form of self‑harm.

I patented an ethical AI wearable to help neurodivergent people manage sensory overload. Turning my S²MHD and DAD Framework research into hardware forced a non‑academic question: when AI is strapped to a human nervous system, who is the data really for, who controls it, and whose comfort does the device ultimately serve?

A large-scale randomized trial in high school math found that unrestricted GPT‑4 help during practice boosted grades in the moment but left students 17% worse off on a later, no‑AI exam. For neurodivergent learners, that’s not just a statistic — it’s a safety warning about how schools are rolling out AI.

Schools, therapy centers, and community organizations are adopting AI tools for autism support without asking the most important question: who built these tools, and who did they ask? On April 17, I'm presenting the answer — and it's not comfortable.

I stood at the stove watching my daughter for 22 minutes while she studied a ceiling fan. She didn't look for me once. That morning became the lens through which I now evaluate every AI tool built for autistic users.

Picture a community where bright lights never blind, alarms never jolt, and meetings pause until every processing speed has caught up. Now imagine Neurodivergent City-States--a place that isn’t a niche clinic or short-lived retreat but an officially recognized jurisdiction—designed, governed, and economically powered by neurodivergent citizens themselves. Radical? Absolutely. Impossible? History says otherwise.