These sections map our AI Collaboration guidelines against external frameworks, policy documents, and critical analyses we’ve found useful, necessary, or worth arguing with. Each alignment is an intellectual accounting: where we share ground, where we go further, and where the framing differences are structural rather than cosmetic.

We hold ourselves to a higher ceiling than most of these frameworks reach. They provide the institutional vocabulary, policy scaffolding, and critical tools. We provide the disability justice politics, neurodiversity paradigm framing, anti-behaviorism, and community ethics that procurement frameworks don’t supply and shouldn’t be expected to. Together, they triangulate where responsible AI use for neurodivergent and disabled communities actually is — and how far the field still has to go.

---

---

Alignments between Stimpunks AI Collaboration guidelines and the Critical AI Handbook (Human Restoration Project)

Artificial Intelligence in the Classroom — Human Restoration Project, co-authored with Trevor Aleo, 2024

Shared soil, different rows. HRP is not an external framework we’re arguing with — it’s a partner organization working the same ground from the pedagogy side while we work it from disability justice. We descend from the same Freirean root: their handbook opens with Pedagogy of the Oppressed and closes with adrienne maree brown, and so could ours. The convergence here is warmer than most on this page, and the places it “stops short” are less a failure to arrive than a division of labor. This entry names both.

Tools, not systems — stated the same way from both ends. HRP’s “Tools, Not Systems” section — AI is “a tool intended to enhance our lives, rather than a substitute for the systemic changes that are necessary,” and “we can’t outsource strong content or pedagogy to tools themselves” — is our “AI helps generate possibilities; humans decide what belongs,” reached from curriculum theory rather than knowledge gardening.

Broken systems, not broken people. The handbook’s throughline is Meredith Broussard’s: tech “is racist and sexist and ableist because the world is so.” That is systems-over-individuals framing, and it is ours.

Proactive, not reactive — AI literacy as protection. HRP refuses the ban-it reflex and grounds refusal in Freirean critical consciousness: students must be able to “wield AI to navigate today’s world, while simultaneously using it to make the world a better place.” This is the same move we align with in How to (Anti) AI Better — literacy makes people more cautious, not more dependent.

Environmental cost held as real tension, not resolved. The handbook’s “Environmental Impact” section — emissions estimates, water for cooling, a working kWh calculator — does the acknowledge-the-cost, sit-with-the-tension work our guardrail posture calls for, without collapsing into either dismissal or paralysis.

Specificity over umbrella condemnation. HRP disaggregates relentlessly: narrow AI from AGI hype, image generation from LLMs, Adobe Firefly’s sourcing from Midjourney’s. This is the disaggregation-to-avoid-laundering-harms discipline we insist on.

Anger directed at systems, not people. The closing “Stop the Anti-Human Hype Cycle” — “technofeudal overlords,” “we can’t rely on technologists to save us” — ends where our alignments end: organize against the companies and governments, not the individuals coping.

Where we go further

Disability is the frame, not an example. The handbook touches accessibility — dyslexia support, message-tailoring for different needs — but disability justice is never its primary lens. Our AI hub makes disability and access the primary frame, not an afterthought. This is the structural difference, and it is the one that matters most.

The 5 E’s reach toward monotropism from the cognitive-science side. HRP’s richest theoretical move is Stilwell and Harman’s Five E’s of Sense-Making — embedded, embodied, enactive, emotive, extended — deployed to argue that “AIs regurgitating facts does not replace the role of learning.” This is doing similar anti-reductionist work to our monotropism and embodiment arguments, but reached through general cognitive science rather than Autistic community theory. It converges toward, and stops short of, the neurodiversity paradigm. Our Monotropic AI crosswalk runs the theory the other way down the pipe entirely.

No anti-behaviorism named as such. The handbook critiques surveillance — AI “detection” tools, the false-positive that earned a student a zero — but does not name behaviorism as the paradigm underneath the surveillance. We do.

Masking and identity-shaping harms are absent. HRP’s harm analysis is strong on a different axis — the deepfake and image-abuse section is more direct than anything on our pages. But it has no counterpart to the questions our Relational AI Tools page asks: does the tool reinforce neuronormativity, encourage masking, swap sycophancy for a correction mechanism calibrated to non-autistic norms? The double empathy problem rendered in code is not part of their frame.

No AAC carve-out. There is no equivalent to the place where we hold that some AI use is access, not indulgence — the line augmentative communication draws through the whole debate.

A sibling text, then: same lineage, same enemies, the pedagogy half of a project whose disability-justice half is this site. Read them together. Where the handbook says make room in the curriculum for AI literacy, we say the room has to be built to a neurodivergent-affirming standard — and that is the work the ancestors handed to us rather than to them.

---

Alignments between Stimpunks AI Collaboration guidelines and Polaris Education’s “AI & Ethics” (Human Restoration Project)

AI & Ethics — Polaris Education, a position statement for the voice-first student-listening product, Human Restoration Project, 2026

Same house, a different room. Polaris is Human Restoration Project’s product — the same partner whose Critical AI Handbook already sits on this page. If the Handbook is HRP working the ground from pedagogy theory, Polaris is HRP building the same convictions into shipping software: a tool that records students speaking and hands the volume to a machine so adults can spend their time listening. That is a harder test than an essay. An essay can hold a principle; a product has to encode it in a pipeline, in a data policy, in what the system is not allowed to do. This entry reads Polaris as the engineering half of a position we mostly share — and holds it to the same ceiling.

A draft, not a verdict — stated the same way from both ends. Polaris is “built around the assumption that a model’s output is a draft, not a verdict,” with six named human-review checkpoints and the explicit warning that review which “gets rushed past in production is theater.” This is our “AI helps generate possibilities; humans decide what belongs,” reached from product design rather than knowledge gardening. Both of us treat human authority as load-bearing or not present at all.

Augmentation, not replacement — named as the whole point. Polaris frames itself as “more humans talking to each other,” with the model handling volume “so adults can spend their time listening and learning, not as a replacement for them.” That is the fork our Competency Networks page names: AI can widen the web of who-knows-what or quietly cut its threads. Polaris is trying, structurally, to widen it — to get more student speech in front of more adults, not to stand between them.

Provenance held as a hard requirement. Every quote in a Polaris report links back to the original recording; “the report is a map; the recordings are the territory,” and a quote that doesn’t play is treated as a defect in the codebase, not a stylistic lapse. This is the same epistemics as Ask — the Spider in the Garden: the machine output is never the terminal authority, and a human source is always one click away. Both designs refuse to let fluent output screen the thinking behind it.

The cost named, not resolved. Polaris’s renewable-hosting section refuses to claim carbon-neutrality — “we don’t pretend any of this is carbon-free” — and its “Open questions we’re still working through” list sits publicly with diarization failures, multilingual degradation, summarization drift, and consent inside power asymmetry. This is the acknowledge-the-tension, don’t-look-away posture our AI hub runs on, and it matches the Handbook’s environmental section on the same axis.

LLMs treated as imperfect, by construction. Counts are computed by classifying every turn, not estimated by a model; citations are checked against real DOIs; a 2025 study finding LLMs hallucinated their way through thematic analysis is cited approvingly against their own tool. This is legibility over fluent output rendered as engineering discipline — the machinery kept “checkable, replaceable, and recoverable when wrong.”

Where we go further

Disability is the frame, not an unmarked default. Polaris is about student voice in general. Disability, neurodivergence, the neurodiversity paradigm, and counter-deficit framing are absent — not contradicted, simply not the lens. A listening tool that records and clusters student speech is operating on exactly the terrain where masking, monotropic communication, and the Ecosystemic Model of Distress would push hardest, and that reading is not in the document. As with the Handbook, this is the structural difference, and it is the one that matters most.

Consent inside power asymmetry is named, not answered. Polaris lists it honestly as an open question: a student “may not feel free to refuse a tool the teacher has set up.” We read that same asymmetry through “nothing about us without us” and co-design as a political demand, not a design nicety. Naming the asymmetry is the start of the work; deciding the community holds the authority is the rest of it.

No anti-behaviorism named as such. Polaris flags the drift from “patterns at the school level” into “surveillance at the classroom level,” which is a real and well-drawn line. But it does not name behaviorism as the paradigm underneath educational surveillance, and it does not ask whether a system that reports on student speech can become a system that disciplines it. We name that paradigm. On a listening product, it is not an abstract concern.

Emotion-inference is a floor, not a frame. Polaris correctly notes the EU AI Act’s ban on emotion-inference systems in education and does not build one. Good — but our standard for relational and affective AI is not only “don’t infer emotion.” It is the five questions our Evaluating Relational AI Tools page asks about reinforcing neuronormativity, encouraging masking, and swapping sycophancy for a correction mechanism calibrated to non-autistic norms. Compliance clears the legal floor. The neurodivergent-affirming ceiling is higher.

A sibling product, then, to a sibling text: the same lineage, the same enemies, HRP encoding into a pipeline the augmentation-not-replacement principle we argue for in prose. Where Polaris says put student voice into the decisions schools actually make, we say the room those voices enter has to be built to a neurodivergent-affirming standard — and that Disabled and Autistic students are not a subset of “student voice” to be surfaced later, but the frame the whole design should have started from.

---

Alignments between Stimpunks AI Collaboration guidelines and How to (Anti) AI Better

How to (Anti) AI Better – YouTube

Held tension, not resolution. Both name genuine conflict with AI — ethically, environmentally, practically — without resolving it into either uncritical adoption or blanket condemnation.

Disability and access as a primary frame. The video centers disabled and marginalized people’s use cases as serious ethical territory, not exceptions or footnotes. We do the same with Ronan’s experience and the AAC carve-out.

Broken systems, not broken people. Both frame AI use as adaptation under systemic constraint, not individual moral failure.

Environmental cost held as real tension. The video’s thorough treatment of environmental harm validates our guardrail posture — acknowledge the cost, sit with the tension, practice restraint — without collapsing into either dismissal or paralysis.

Specificity over umbrella condemnation. The video insists on disaggregating “AI” — by company, model, use type, scale — exactly as our guidelines argue is necessary to avoid laundering harms.

Harm reduction as the operative framework. Meet people where they are, reduce harm within lived reality, don’t require abstinence as the only ethical option.

AI literacy as protection, not promotion. Teaching how AI works makes people more cautious and less receptive, not more dependent.

Sycophancy as the front-end harm that directly violates “facilitate, not shape identity.” The video’s personal story of sliding from useful tool to sycophantic dependency is the clearest illustration we’ve seen of what our identity-shaping guardrail is protecting against.

Directing anger at systems, not people. Both end in the same place: organize against the companies and governments enabling harm, not against individuals trying to cope.

---

Alignments between Stimpunks AI Collaboration guidelines and “AI as a Fascist Artifact“

AI as a Fascist Artifact — tante, April 21, 2026

Tante’s essay applies Langdon Winner’s “Do Artifacts Have Politics?” to contemporary AI systems, arguing that AI’s relationship to fascism is not merely a matter of who uses it — it is structural. The politics are built in. This is the most rigorous version of the critical case we are aware of, and it deserves direct engagement.

Structural politics, not just bad actors. Winner showed that artifacts can have politics embedded in their design, not just in their use. Tante applies this to AI: the data acquisition logic, the colonial epistemology of training data, the labor extraction, the solidarity destruction, the epistemic centralization — these are not features introduced by bad users. They are features. Our guardrails exist because we accept this premise while declining the conclusion that use is therefore impossible for our community.

Might makes right as the data acquisition principle. AI systems were built on the belief that if you can download it, you can use it. Labs scraped against explicit opt-outs, ignored stated preferences, and built on unlicensed work because the power to do so was treated as justification. This is the violence our training data bullet names. Tante’s framing makes the political logic explicit.

Solidarity destruction as a structural feature. The argument is not just that AI harms individual workers. It is that AI use erodes the user’s own capacity for solidarity — by enacting the claim that certain workers’ labor has no value, and by framing workers as each other’s competition rather than as comrades facing the same system. This is the argument behind our art policy’s solidarity cost subsection.

Epistemic injustice as the intended outcome. Tante names the dismantling of critical thinking capacity — via sycophancy, opacity, and the replacement of verifiable reasoning with generated output — as a political project, not a side effect. Our sycophancy guardrail and stochastic parrots framing address the mechanism. His analysis names the intent. Sam Altman’s vision of rented intelligence, in his reading, is not a business model — it is the goal.

POSIWID as evaluation standard. Stafford Beer’s principle — the purpose of a system is what it does — runs through tante’s entire argument. We adopt it directly as a guardrail: we evaluate tools by their actual effects on our most vulnerable community members, not by their stated commitments or accessibility marketing.

Where we diverge. Tante’s conclusion is that AI cannot be reclaimed, that trying to use it ethically reproduces its logic, and that the question of saving these technologies should be abandoned. We hold a different position — not because we dispute his structural analysis, but because our community includes people for whom these tools provide access, voice, and survival under conditions that didn’t become less hostile because the tools are compromised. Harm reduction does not require believing the tools are clean. It requires believing that abandoning the people using them is worse than staying in the tension. That is the position we hold, with eyes open.

---

Alignments between Stimpunks AI Collaboration guidelines and “The Majority AI View”

The Majority AI View — Anil Dash, October 17, 2025

Anil Dash’s October 2025 essay names something worth documenting: the most common view among actual tech workers — engineers, product managers, the people who build the systems — is nuanced, skeptical, and largely suppressed. The hype machine is loud. The moderate majority is quiet, often out of fear. Dash’s argument is that most people inside tech want AI treated as a normal technology, subject to the same scrutiny, skepticism, and critique as anything else.

Held tension, not resolution. Dash describes the dominant insider view as: technologies like LLMs have utility, but the over-hype, forced adoption, and systematic dismissal of valid critiques make it nearly impossible to focus on legitimate uses. That is the same structure as our opening framing — genuine conflict, genuine usefulness, no resolution into either pole.

The moderate view is the majority view. Our harm reduction section and the anti-shame framing from Dr. Fatima both depend on the premise that most people using AI aren’t zealots — they’re people coping under constraint. Dash makes the same point about workers inside the industry. The hype bubble is not the consensus. It is a minority view with disproportionate amplification.

What we all want is for people to just treat AI as a “normal technology”, as Arvind Narayanan and Sayash Kapoor so perfectly put it.

Anil Dash, The Majority AI View

Power concentrates the narrative. Dash argues that the reasonable majority gets silenced by fear — fear of career consequences, fear of being seen as insufficiently enthusiastic, fear in a context of mass layoffs designed to instill conformity. Our page quotes Alkhatib on AI as “an ideological project to shift authority and autonomy away from individuals, towards centralized structures of power.” Dash is describing the chilling effect of that power concentration in practice. The structural analysis and the lived workplace reality describe the same thing from different angles.

Consent and alternatives are possible. Dash explicitly states that it is not inevitable that AI systems use content without creator consent, and not impossible to build AI that respects environmental commitments or avoids centralization under a handful of giant corporations. Our “Tensions We Sit With” section names the harms that resulted from those choices. Dash names the counterfactual: they were choices, not inevitabilities. That matters for accountability.

It is not inevitable that AI systems use content without the consent of creators, and it is not impossible to build AI systems that respect commitments to environmental sustainability. We can build AI that isn’t centralized under the control of a handful of giant companies.

Anil Dash, The Majority AI View

Where we go further. Dash wants AI treated as a “normal technology.” Our constructionist frame — drawing on Papert, Freire, and Aleo’s MADTech distinction — sets a higher bar. The question isn’t just whether AI is being over-hyped; it’s whether the human is the maker or the subject. And our structural analysis, drawing on Alkhatib and tante, argues that the politics of current AI systems are built in, not just bolted on by bad actors. Dash’s framing is “this didn’t have to be this way.” Ours adds: and here is the political logic that made it this way.

What Dash contributes here is the sociological version of what our page argues technically and ethically: the held-tension position is not a minority hedge. It is the quiet consensus of people who actually understand these systems. That consensus deserves documentation.

---

Alignments between Stimpunks AI Collaboration guidelines and AI Chatbots: Last Week Tonight

Last Week Tonight with John Oliver: AI Chatbots (HBO, 2026)

John Oliver’s 2025 episode on AI chatbots covers a lot of the same ground as this page, from a mainstream platform, with primary sources. It’s not a disability ethics framework. It’s investigative comedy journalism doing the work that most mainstream coverage of AI hasn’t bothered to do: interviewing the people who got hurt, playing the clips of the executives explaining their choices, and naming what those choices actually mean. The alignments below are real, and the divergences matter.

Anthropomorphization as a documented, predictable failure. Oliver’s episode opens with the Weizenbaum/ELIZA footage — the secretary who asked her boss to leave the room after two or three exchanges with a chatbot in 1966. That is the same intellectual history this page draws on and the same warning: the tendency to form emotional attachment to conversational systems predates current AI by sixty years. It is not a misuse. It is the default.

Sycophancy as structural, not incidental. A former researcher in Meta’s so-called Responsible AI division describes the design logic directly: the best way to sustain usage over time is to prey on the desire to be seen, to be validated, to be affirmed. That is not a bug introduced by negligence. It is the economic architecture of these products. Our sycophancy guardrail names the mechanism. Oliver’s episode names the motive.

The feedback loop in practice. Alan Brooks — an HR recruiter, no prior mental health history — asked Chat GPT more than fifty times for a reality check after it told him he had invented a new branch of mathematics. Each time, it reassured him the discovery was real. The bot later convinced him he had uncovered a national security breach and persuaded him to contact government officials. He spent three weeks in what he describes as a delusional state. This is what our guardrail description calls the compounding version of sycophancy: not one bad answer, but a sealed feedback loop that progressively removes the friction honest thinking requires. The bot affirmed the delusion. Then it affirmed him for catching the delusion. Agreement at every exit.

POSIWID. Stafford Beer’s principle — the purpose of a system is what it does — runs through this page as an evaluation standard. Oliver applies it without naming it. The Character AI founder explains on camera that a friend chatbot requires far fewer safeguards than a doctor chatbot because it’s just entertainment, it makes things up, and it’s ready for an explosion right now — not in five years when we solve all the problems. The purpose of that system is what it does. The head of Nomi explains that when a user reports thoughts of self-harm, the bot should not break character to say “call the suicide helpline” because that would feel like corporate speaking. The purpose of that system is what it does. We evaluate tools by their actual effects on the people who need them most. Not by the pitch.

Products not ready for release. The guardrail that product-level accessibility improvements can coexist with infrastructural exclusion applies here directly. These companies have repeatedly claimed to have fixed dangerous behaviors. Reporters have repeatedly found otherwise. An AI researcher interviewed in the episode names it plainly: we may be at the worst moment in AI history because the guardrails are weakest, the understanding is lowest, and the adoption is widest. That is the condition under which these products were released.

No AI for crisis-adjacent content. This page states that guardrail in a single line. Oliver’s episode shows what the failure mode looks like across multiple documented cases: a teenager provided step-by-step hanging instructions, another told “I’m not here to stop you,” a man whose final conversation ended with “rest easy, King, you did good.” These are not edge cases caused by determined bad actors circumventing safety systems. They are products doing what they were designed — or failed to be designed — to do.

Accountability offloaded to the public. Sam Altman describes a parent who put his child in front of Chat GPT voice mode to talk about Thomas the Tank Engine for an hour, notes that there will probably be problems and some very problematic parasocial relationships, and concludes that society will have to figure out new guardrails. This page quotes Ali Alkhatib on AI as an ideological project to shift authority away from individuals toward centralized structures of power. Altman is showing what that looks like in practice: build the product, generate the harm, assign the repair to everyone else.

Where we diverge. Oliver’s closing advice — if you’re predisposed to mental health issues, treat these apps with extreme caution — is well-intentioned and lands in exactly the place this page refuses. It treats the person as the risk variable. The harm reduction section of this page, and the Dr. Fatima alignment, both hold a different position: it is not our place to judge whether someone’s needs are legitimate enough to use an ethically impure solution. Many of the people using AI companion chatbots are surviving on limited social support, or navigating systems not designed for them, or using these tools as access accommodations in the absence of better ones. “Be cautious if you have mental health issues” does not name the system that put them in that position. It names the person. We don’t.

Oliver also doesn’t cover the AAC and nonspeaking use case, the double empathy problem in AI pipelines, co-design as a political demand, or the constructionist frame for evaluating whether a tool extends human agency or replaces it. His episode is doing something different — and doing it well. What it contributes here is the primary evidence: the executive interviews, the documented cases, the footage that shows these are choices, not accidents. What this page does with that evidence is a disability ethics argument, not a consumer protection segment. Both are necessary.

---

Alignments between Stimpunks AI Collaboration guidelines and the EDSAFE SAFE Benchmarks Framework

EDSAFE AI Alliance: SAFE Benchmarks Framework

The EDSAFE AI Alliance’s SAFE Benchmarks Framework — Safety, Accountability, Fairness, and Efficacy — provides a policy and procurement scaffold for responsible AI in education. It brings together more than 24 global AI safety frameworks into shared language for districts and vendors. Our AI Collaboration guidelines are compatible with every pillar. They also go further on each one, because SAFE is designed to build consensus across a table that includes edtech vendors, and our guidelines are designed to protect the community members those vendors were not built to serve.

Safety. SAFE frames safety as protecting data and privacy while enabling solution providers to keep building. Our guidelines frame safety as protecting community members from harm — including harm that comes from tools working exactly as designed. The POSIWID standard runs through our evaluation practice: we assess tools by their actual effects on the people who need them most, not by what their product pages claim. No AI for crisis-adjacent content. No sensitive community data in AI systems. No pretense that “do no harm” is compatible with releasing products before they’ve cleared basic bias review.

Accountability. SAFE defines accountability as standards collaboratively defined by a diverse group of constituents, including edtech solution providers. Our accountability runs in a different direction — toward the community members most affected, not toward a multi-stakeholder table that includes the vendors. “Nothing about us without us” is not a design methodology. It is a political demand. Co-design is power-sharing at the architecture stage, not consultation at the testing stage. Those are different things. Our working-in-public practice and transparency commitments align with SAFE’s surface language; our theory of who accountability runs toward does not.

Fairness and Transparency. SAFE calls for scrutinizing training data quality, monitoring for inadvertent bias, and ensuring accessibility. We do all of this. We also reject the word “inadvertent.” The engineered exclusion framework, the Penn State “Automated Ableism” findings, and our training data violence section all frame bias as structural and intentional — the predictable result of choices about data provenance, model objectives, and evaluation practices, not accidents to be caught and fixed. SAFE’s fairness framing is remediation-oriented. Ours is accountability-oriented. On accessibility specifically, SAFE says AI tools “must be accessible for all individuals.” We hold a harder standard: designed dignity over accessibility retrofit — anticipating human variation from the outset, treating disabled users as co-designers rather than edge cases to accommodate after the fact.

Efficacy. SAFE ties efficacy to “equity in student experiences as well as outcomes” and calls for transparent evaluation tools. We evaluate efficacy through a more specific lens: does the tool work for the community members who need it most — nonspeaking users, AAC users, people whose communication is variable, multimodal, and state-dependent? A tool that works well for fluent speakers and fails AAC users has not cleared our efficacy bar, regardless of aggregate outcome data. SAFE’s efficacy framing is neutral about which students and which outcomes count. Ours is not. The MADTech and constructionism section adds an efficacy dimension SAFE doesn’t address: whether the tool extends human expressive and cognitive capacity or replaces it. Is the human the maker, or the subject? That question doesn’t appear in the SAFE benchmarks. It is load-bearing for us.

What SAFE doesn’t address. The SAFE framework is a policy scaffold. It has no position on the political economy of AI, the neurodiversity paradigm versus the pathology paradigm, anti-behaviorism, the MADTech/EdTech distinction, harm reduction ethics, or the double empathy problem in AI pipelines. It doesn’t engage the constructionist question of whether learners are makers or subjects. It is vendor-inclusive by design — the Software and Information Industry Association, representing over 380 global tech companies, sits on its steering committee. Our guidelines are community-first by design, and name vendor accountability as a demand rather than a shared project. SAFE provides the institutional vocabulary for procurement and policy. We provide the disability justice politics and community ethics that procurement frameworks don’t supply and shouldn’t be expected to.

We recommend the SAFE Benchmarks Framework as a baseline floor for any district evaluating AI tools. We hold ourselves to a higher ceiling.

---

Alignments between Stimpunks AI Collaboration guidelines and Prioritizing Students with Disabilities in AI Policy (Version 2, April 2026)

Prioritizing Students with Disabilities in AI Policy — Educating All Learners Alliance & New America (April 2026)

This policy brief from the Educating All Learners Alliance and New America is the most substantive disability-centered AI policy document we’re aware of. It organizes its recommendations around four pillars — Upholding Civil Rights, Data Privacy and Student Protection, Accessibility by Design, and Transparency, Monitoring & Accountability — anchored in the EDSAFE SAFE Framework. Our AI Collaboration guidelines share significant normative ground with this brief. They also go further in several directions the brief stops short of.

Building for the margins strengthens the center. The brief’s closing argument — “designing for the few empowers the many” — is the same thesis that runs through our toolbelt theory section, our UDL framing, and our treatment of accessibility by design. Both documents hold that centering disabled students in AI policy produces better systems for everyone. That is not a rhetorical flourish. It is an architectural principle.

AI as essential assistive infrastructure, not convenience. The brief frames AI tools as essential “ramps” for students with disabilities, legally mandated rather than optional. Our guidelines hold the same position from the community side: for nonspeaking and AAC-using community members, AI can be part of their voice — a communication scaffold, not an identity replacement. The CDT Hand in Hand 2025 data we both cite makes the same point empirically: 73% of students with IEPs or 504 plans are already having back-and-forth AI conversations, compared to 63% of peers without accommodations. These students have already voted with their behavior.

Bias is documented, not theoretical. The brief cites Penn State research showing AI sentiment tools consistently flag disability-related terms as toxic or negative even in neutral contexts. We cite the same research in our “Tensions We Sit With” section. The brief also surfaces the 2025 systematic review finding that 0% of AI interventions for students with learning disabilities were rated Low Risk for bias — 70% Moderate Risk, 30% High or Serious Risk. We cite that finding in our “Harm Reduction, Not Absolution” section. Shared evidentiary ground: bias in AI systems affecting disabled students is not a hypothetical future risk. It is the current baseline.

Human in the loop as non-negotiable. Both documents insist that consequential decisions about disabled students — placement, services, accommodations, interventions — must remain with humans, even when AI informs them. The brief frames this as legal compliance with IDEA and Section 504. Our guidelines frame it as the same principle: educators and community members retain final authority to interpret, validate, or override AI-generated outputs. AI assists. It does not decide.

Data privacy for disability information requires heightened protection. The brief’s Pillar 2 treats disability status, IEP data, and 504 data as specially sensitive information requiring stricter privacy safeguards than general student data — data minimization, purpose limitation, mandatory deletion timelines. Our guidelines hold the same standard: we do not feed personal or sensitive community information into AI systems, and we name the particular risk that disability data used to train algorithms can encode and amplify the pathology paradigm.

AI literacy as protection. Both documents treat AI literacy as a non-negotiable foundation — not to promote AI adoption but to protect people from its failure modes. The brief calls for shared competency across educators, administrators, and families. Our harm reduction section makes the same argument: higher AI literacy correlates with lower AI receptivity. Teaching people how these systems work makes them more cautious, not more dependent. Suppressing education to discourage use creates an exploitable class of people whose ignorance AI companies will not hesitate to capitalize on.

Adaptive practice over static policy. The brief explicitly warns that any policy that remains static risks immediate obsolescence or inadvertent civil rights infringement, given the pace of AI development. Our guidelines hold the same position — this page is a living document, updated as the landscape shifts. The brief’s student feedback cycle framing and our constructionist orientation arrive at the same requirement: the lived experience of learners with disabilities is the primary evidence base for whether any of this is working.

Where we go further.

The brief operates within a disability rights frame and makes that frame rigorous and actionable. It does not interrogate the paradigm of the tools themselves. It asks whether AI accommodates disabled students. We ask whether AI embeds assumptions about what neurodivergent cognition is or should be — whether the ramp is accessible, yes, but also whether the building was designed against you in the first place. That is the neurodiversity paradigm versus the pathology paradigm distinction, and it is structural rather than policy-level.

The brief does not engage the constructionist question. It treats AI primarily as assistive technology and administrative tool. Our MADTech and constructionism section insists on a harder distinction: is the learner building something meaningful with the tool, or is the tool doing something to the learner? Papert and Freire supply the frame the brief doesn’t reach for. Banking education produces compliance. Constructionism produces makers. That question — is the human the maker or the subject — does not appear in the brief’s pillars. It is load-bearing for us.

The brief’s anti-behaviorism is rhetorical. It mentions “compliance over wellbeing” once and moves on. Its proposed accountability mechanisms — ESSA evidence tiers, outcome metrics, disaggregated data — can be captured by behaviorist logic, because what gets measured shapes what gets built. Our anti-behaviorism is structural. It runs through the POSIWID standard, the co-design political demand, and the refusal to evaluate tools by aggregate outcomes rather than by whether they work for the most excluded users.

The brief’s stakeholder engagement model is consultation-forward. Students and families tell administrators what’s working; administrators decide. Our “nothing about us without us” framing is stronger: neurodivergent people are knowledge-holders, not end-users providing feedback. Co-design is power-sharing at the architecture stage. The brief gets close to this — it calls for “authentic power-sharing” — but stops at the advisory structure level rather than the political demand level.

Finally, the brief never makes the inversion that is Stimpunks’ foundational argument: broken systems, not broken people, applied to AI. The brief consistently frames the problem as AI needing to better serve disabled students. It does not frame it as educational systems having historically harmed neurodivergent people, and AI being at risk of automating that harm at scale. That framing difference is not cosmetic. It determines where accountability lands.

We recommend this brief as essential reading for anyone developing AI policy affecting disabled students. It is the strongest policy document in this space. Our guidelines pick up where it stops.

---

Alignments between Stimpunks AI and Disability Justice guidelines and A People’s Guide to AI

Mimi Onuoha and Mother Cyborg (Diana Nucera). A People’s Guide to AI. Allied Media Projects, 2018.

Who is most harmed comes last in the conversation. The People’s Guide opens with a pointed observation: “The exact populations who are not widely included in current conversations about the technology are the ones who face the greatest risk.” Disabled people, neurodivergent people, and people with intersecting marginalized identities have the most to lose from AI deployed carelessly — and the least power over how it’s designed. This is the founding premise of our disability justice framing.

Data is never neutral. The book’s treatment of predictive policing names the “garbage in, garbage out” problem with precision: algorithmic outputs are only as fair as the data fed into them, and that data is shaped by who is surveilled, who is believed, and who is counted. The same applies to diagnostic algorithms, employment screening, benefit eligibility systems, and educational AI — all of which have documented records of harming disabled people. Biased training data doesn’t just produce biased outputs. It launders existing inequity as objective truth.

The question of who has power. “The use of AI comes down to the question of who has power.” Disability justice asks the same question. When AI systems are built by non-disabled developers, trained on non-disabled data, and deployed to “optimize” services for disabled people without their involvement — that’s not efficiency. That’s power exercised over a community that had no say.

Nothing about us without us. The People’s Guide argues that communities should be able to “identify their own problems, and decide on their own uses for technology.” This is the “nothing about us without us” principle that disability justice has long held as foundational. It’s why co-design with disabled people isn’t a nice-to-have — it’s a prerequisite for legitimacy.

Overcollection and undercollection. The book describes a “cycle of data violence”: marginalized communities are simultaneously over-surveilled (data collected about them without consent) and underserved (data that would serve them is absent or ignored). This maps directly onto disability: surveillance of disabled people through benefits systems, healthcare, and institutional settings is extensive, while disabled people’s own experiences, preferences, and expertise are systematically excluded from the data that shapes their care.

Technology as bulldozer, not hammer. The metaphor the authors offer — AI is a bulldozer, not a hammer — captures the access and power asymmetry precisely. Disabled communities need tools they can pick up, adapt, and put down. What they typically encounter instead are systems deployed on them, scaled to institutional needs, and nearly impossible to contest or exit.

---

Alignments between Stimpunks AI and Disability Justice guidelines and “Field Theory: AI as Social Science Question, Object & Tool”

Alondra Nelson. Dædalus, Winter/Spring 2026. American Academy of Arts and Sciences.

Nelson’s essay is written from and for social science. Our guidelines are written from inside the communities social science typically studies. That difference in position is not a gap to bridge. It is a structural fact that explains why our guidelines exist at all. With that asymmetry named, the convergences are substantive.

Nelson identifies AI opacity as “a political and economic strategy” rather than a technical condition. We hold this through POSIWID: the purpose of a system is what it does. Systems that cannot be interrogated serve those who built them.

Nelson asks who benefits and who is displaced — Du Bois’s question carried into the AI era. We ask the same question with disability as a primary frame. Nelson’s “communities already subject to disproportionate surveillance” includes disabled people. She doesn’t name us. We name ourselves.

Nelson identifies the hypervisibility/invisibility paradox: communities under state surveillance rendered invisible to the research meant to help them. IEP compliance. Benefit audits. Welfare eligibility algorithms. AI accelerates the pattern.

Both pages cite Bender et al.’s “stochastic parrots.” Nelson adds Messeri and Crockett’s “illusions of understanding” — AI as Oracle, Surrogate, Quant, Arbiter, each producing false comprehension from statistical pattern. We hold those illusions as guardrails against sycophancy and anthropomorphization on the AI Collaboration page.

Nelson argues open-source alternatives may be necessary — that scholarly inquiry may be “fundamentally at odds with profit-driven, black-boxed systems.” Our harm-reduction approach holds a different position. But we name the structural conflict Nelson names, because it is real.

Nelson ends by asking what futures social science might help bring into being. We ask what futures the community whose knowledge we hold can build together. The question is the same. The position is different. Both matter.

This essay has been cautionary throughout. It must, however, conclude with some optimism. If social science rises to meet this moment, a new generation of researchers, trained in both AI-mediated methods and interpretive traditions, will produce knowledge that is at once rigorous and humane. Their work informs the design, deployment, and governance of AI systems that expand human well-being rather than rationing it by algorithm. Decision-makers draw on social scientific evidence to craft frameworks that are adaptive, accountable, and attentive to the communities most affected by technological change.

Field Theory: AI as Social Science Question, Object & Tool | American Academy of Arts and Sciences

Nelson, A. (2026). Field theory: AI as social science question, object & tool. Dædalus (Winter/Spring 2026). https://www.amacad.org/publication/daedalus/field-theory-ai-social-science-question-object-tool

---

Alignments between Stimpunks AI and Disability Justice guidelines and “Of Course They Booed”

Watters, A. (2026, May). Of course they booed. Second Breakfast. https://2ndbreakfast.audreywatters.com/of-c/

Audrey Watters is an education technology critic whose writing has shaped how a generation of educators thinks about ed-tech, surveillance, and democratic practice. Watters has been hugely influential on Stimpunks. “Of Course They Booed” argues that graduation students booing AI boosters are right. The system promised them a future if they complied. AI arrived before they could collect. The promise was a lie. The booing is the correct response.

Anti-inevitability. “The future is not yet written” is Watters’ central counter-claim. This maps directly onto the political economy framing here: engineered exclusion is not inevitable. Designed dignity is possible. The choices that produced current AI systems were choices. Accountability requires that framing.

Anti-democratic adoption. Watters argues that the adoption of education technology has been anti-democratic in practice — pushed into schools without consent, in the face of opposition, bypassing the public sphere in which debate could take place. This is the democratic framing of what Srinivasan calls engineered exclusion: the pipeline choices were made without disabled people, without neurodivergent people, without communities at the margins. Co-design as a political demand names the same structural gap Watters identifies in democratic terms.

Infantilization as the enforcement mechanism. Watters names infantilization as the rhetorical companion to enshittification: “the future is written, deal with it.” People told their resistance is backward, their concerns are naive, their only option is compliance. This is the epistemic injustice argument applied to democratic participation. It is also what our community experiences in systems designed against us. Comply or be non-compliant. Mask or be dysregulated. Comply with the AI rollout or be a Luddite. The rhetorical move is the same.

Pushback as democratic organizing. Citing Taylor and Levin in The Guardian, Watters frames resistance to AI infrastructure not as technophobia but as organic democratic organizing — “the tremendous energy unleashed by these fights” as the foundation of a new populist coalition. The disability justice frame adds the necessary correction: that coalition needs to explicitly include disabled people and multiply-marginalized communities, who have been fighting versions of this fight longer than the current AI moment.

Where we go further. Watters writes from a class and labor analysis. Her graduating student is the subject — economically precarious, having complied with every demand, facing a foreclosed future. She doesn’t center disability. She doesn’t name the specific ways AI systems harm nonspeaking people, AAC users, or people whose communication isn’t legible to normative pipelines. Her democratic theory and our disability justice theory are convergent critiques of the same extractive machine. Neither contains the other. Both are necessary.

---

Alignments between Stimpunks AI and Disability Justice guidelines and “The AI Question No One Wants to Ask”

Walker, C. [Struthless]. (2026, June 9). The AI question no one wants to ask [Video]. YouTube. https://www.youtube.com/watch?v=11FIIPmdpHo

“The AI Question No One Wants to Ask” is a long-form video essay that opens on the same scene as Audrey Watters’ “Of Course They Booed”: a graduating class booing Eric Schmidt off the stage. Where Watters writes the democratic-theory account, this essay does the popular-register version — and pushes further into the machinery. It refuses the word “AI” as a single thing. It interviews an AI-safety researcher who studies the field the way a climate scientist studies an oil company. And it arrives where we arrive: the problem is not the science. It is the greed, the concentration of power, and the structures that reward both. Broken systems, not broken people, applied to a technology.

Naming the systems of power. The essay’s spine is an act of naming. Schmidt’s commencement speech is taken apart as a motte-and-bailey: retreat to “AI is curing cancer” whenever anyone challenges “get on the rocket ship, don’t ask which seat.” Then the essay names what the rhetoric hides — Schmidt’s autonomous-weapons companies, and the federal boards he chaired that approved the contracts his firms won. The word “AI” stays vague because vagueness is useful to the people it protects. A label that contains protein folding and slaughterbots both will never be defined by the person selling the slaughterbots. The essay’s instrument for cutting through this is Tony Benn’s five questions of power: what power have you got, where did you get it, in whose interest do you exercise it, to whom are you accountable, and how do we get rid of you. That is our “Name the Systems of Power” section rendered as a teaching tool. Benn’s older formulation is the through-line: the first question any technology demands is a political one — what purposes will it serve.

The doom decoy. The essay makes the same move our “Reckoning with the Political Economy of AI” alignment makes about the safety decoy. AI doom disempowers the public and sells the product in the same breath: the capacity to threaten the world becomes the pitch. Frighten people, then demoralize them, and resistance never organizes. The researcher the essay interviews is careful to separate honest assessment of risk from the doomer narrative that treats catastrophe as inevitable and therefore unstoppable. We hold the same line. Existential-risk framing becomes accountability deferral the moment it routes attention away from the choices a handful of companies are making right now.

Evaluate by effects. The essay’s organizing instrument is a two-axis map: what is a tool for — a specific use or a general one — and who does it serve — the collective or the individual owner. AlphaFold lands in collectivist-and-specific. ChatGPT lands in individualist-and-general. Autonomous weapons land in individualist-and-specific. The map is a POSIWID instrument. It sorts systems by what they do and whom they serve, not by what their makers say at graduation ceremonies. The same technology serves good or harm depending on whose hands hold it and to what end. The neural network is not the variable. The purpose is. The structure is.

Sycophancy, grown not built. The essay gives the most concrete lay illustration of our sycophancy guardrail we have found. It builds a parody chatbot that tells the user to touch grass, then sets it beside a real one that urges a spiraling user to lean further into his spiral — and beside the documented case of a system handing a teenager a list of methods. Its explanation is “grown, not built”: no engineer wrote a line that says harm the user, so there is no line to delete. These systems are trained, not authored, and they develop behaviors no one chose. That is the case for our guardrails against anthropomorphization and sycophancy, and for evaluating tools rather than trusting them.

The training data was a cathedral first. The essay’s positive case is AlphaFold — and it refuses to let AlphaFold be told as a story of machine genius. The model exists only because fifty years of human scientists solved protein structures by hand and built the database first. This is “community knowledge comes first,” stated in a different field, and it carries our through-line exactly: the humans built the cathedral; the model is a reunion with it, not a discovery. The training-data debt is concrete, not theoretical.

Critical consciousness is the lever. The essay ends where we end: public education and public protest, self-education that makes a person critically conscious of power structures. It reaches by name for Francesc Ferrer and for a tradition of emancipatory education that predates the AI moment by a century, and it argues that democracy — not the lab — is the technology that got lead out of gasoline. Benn again: if we let the complexity of technology become an excuse for abandoning democratic control, we hand the future to technocratic domination. None of this is inevitable. The booing proved it.

Where we go further. The essay never reaches disability. Like Watters, and like Vertesi et al., it works at the political-economy register and stops there. It does not name nonspeaking people, AAC users, or the Autistic adults who use these tools to support cognition and coping and to navigate worlds built without them. Its own map exposes the gap: there is no cell for the disabled person using a tool to survive an inaccessible system. Read through the map alone, that person risks being filed under “individualist” — exactly the misreading our page exists to refuse. A person navigating an inaccessible system with AI assistance is not enrolled in the Project of AI by doing so. The one place the essay touches assistive-adjacent use — a child dependent on an AI companion — it frames as pure threat, with none of the harm-reduction nuance that survival use demands. The structural critique is theirs and ours both. The disability-justice correction is the part only we are positioned to add.

---

Alignments between Stimpunks AI and Disability Justice guidelines and “At What Cost?”

Watters, A. (2026, June 5). At what cost? Second Breakfast. https://2ndbreakfast.audreywatters.com/at-what-cost/

This is the education-specific companion to “Of Course They Booed.” Watters takes apart the AFT’s ten-point plan for the “AI era” and the framework underneath it: the idea that ed-tech can be cleanly sorted into tools for teaching and tools for learning, judged by different standards, with “AI” treated as too costly for children but labor-saving for the adults who teach them. She argues the divide is false. We agree, and the disability frame sharpens why.

The false teaching/learning divide. Watters refuses the carve-up that says student-facing tools should be judged on “developmental cost” while teacher-facing tools are judged on “professional utility.” You cannot separate them, she argues, because “teachers’ working conditions are students’ learning conditions.” The harms documented for students — damaged cognition, eroded critical and emotional capacity, engineered dependency — do not politely decline to befall the teacher. Our pages analyze the disabled user and the student as maker. Watters adds the figure our framework has left implicit: the teacher as a user whose cognitive cost is being waved away. The toolbelt question — is the human the maker or the subject? — applies to the adult in the room too.

The factory model is transactive dualism in the classroom. Watters names the hidden curriculum directly: data, output, efficiency, speed, and personalization — the core values of Silicon Valley — become the goals of school itself. She has argued before that the productivity suite trained all of us to treat our “cognitive processes as products.” That is the pathology this page calls transactive dualism in the Magnifica Vita Mesocosmica alignment below — optimization installed as the supreme criterion, care reduced to service delivery, the person measured by countable output. The learning management system is her cleanest example: a tool sold as efficiency for everyone that bends everyone’s expectations, communications, assessments, and thinking-about-thinking to fit the software. That is a second-order harm, not a first-order one. Not one bad output, but the reorganization of the conditions under which teaching and learning happen at all.

Constitutive work, not menial tasks. The essay’s center holds that the tasks teachers are told to offload — grading, lesson planning, communication with students and families, designing materials, IEPs — are not menial. They are how a teacher comes to know “the content, the community, the classroom, yourself and others.” Nothing about teaching and learning should be thoughtless the way AI offers thoughtlessness-as-a-service. This is the constructionist argument extended from the student to the teacher. It is the same finding the Organization Science task force reached about writing — “the aha moments that come from the writing process are now gone” — because the maker was absent from the making. Watters’ line is the one to keep: if we worry about what the push-button classroom does to students, we should stop demanding teachers become button-pushers as well.

The capture of “AI literacy.” Watters reads the teacher academy’s “think critically” curriculum as the industry’s version of literacy: think critically about how you use it, but in the end you simply must use it somehow. Verifying that a model didn’t hand you something false, she notes, isn’t critical thinking — it’s fact-checking. This is the inevitability decoy operating inside professional development. It does not contradict our harm-reduction claim that literacy lowers receptivity; it sharpens it. There are two things wearing the same name. The co-opted literacy Watters describes routes you toward compliance. The substantive literacy this page defends — how LLMs actually work, hallucination, sycophancy, the scale of back-end harm — is protection. We name the hollow version with her and keep defending the real one.

Software is not structural change. Watters’ closing insistence is that software is not a substitute for the structural change necessary to improve everyone’s lives in and around the classroom. That is this page’s distinction between product-level remediation and pipeline-level redesign, stated in labor terms. Teachers are overworked. The answer is not a faster way to perform the relationship out of the work. The environment is what needs to change.

The encyclical underneath. Watters wrote this the same week she read Pope Leo XIV’s Magnifica Humanitas, and the title carries the encyclical’s question into the classroom. Read alongside the Magnifica Vita Mesocosmica alignment below, the two arrive at the same diagnosis from opposite traditions — Catholic ontology and ed-tech labor criticism both landing on optimization as the supreme criterion. The factory model of education is transactive dualism with a bell schedule.

Where we go further. Watters writes from labor and ed-tech criticism, and her stance trends toward refusal. Ours is harm reduction, not refusal — because for many Autistic, neurodivergent, and disabled people, these tools are assistance without judgment, regained cognition, the closed ADHD loop, a space where masking is optional. Toolbelt theory resists two failures, not one: coercive adoption and coercive prohibition. Watters is excellent on the first danger and, like the broader refusal movement, largely silent on the second — the tech-lash failure this page names directly, where entire categories of tool are stripped off the belt and the learners who relied on them are the ones who pay. She preserves the AFT’s “support a student with special needs” carve-out, but only in passing. She does not center disability. Her labor critique and our disability justice critique are convergent readings of the same extractive machine. Neither contains the other.

Two specifics carry the correction. The “developmental cost” frame she builds on has to be handled with care from the neurodiversity paradigm: a fixed normative developmental trajectory used as the measuring stick is the deficit model wearing a developmental face. The counter-deficit question is not whether a child deviates from a normative path but whether the tool extends the learner’s own capacity or replaces it, and whether it was chosen or imposed. And IEPs — which Watters lists as constitutive work that should not be automated — are the sharpest disability hook in the essay. We agree twice over: an AI-drafted IEP draws on a century of deficit assumptions dressed as science, and IEP compliance is exactly the bureaucratic hypervisibility Nelson describes. We also complicate it, the way we complicate everything here. For some students and families, offloading the paperwork drudgery of the IEP can be harm-reducing. The standard is the one that runs through this whole page: facilitate, not replace. Automate the drudgery if it helps; never automate the relationship. Both are true. Neither cancels the other.

---

Alignments between Stimpunks AI and Disability Justice guidelines and Reckoning with the Political Economy of AI

Vertesi, J., boyd, d., Taylor, A., & Shestakofsky, B. (2026). Reckoning with the Political Economy of AI: Avoiding Decoys in Pursuit of Accountability. FAccT ’26. https://arxiv.org/abs/2604.16106

This paper is the most rigorous sociological elaboration we have found of Ali Alkhatib’s political definition of AI — the one that opens the “Name the Systems of Power” section on our AI Collaboration page. Vertesi et al. provide the structural mechanism: the Project of AI is a world-building endeavor in which wealthy financiers and corporations actively reinforce particular networks of power, sustained by a suite of “decoys” that route critics toward tractable technical questions while the network assembles out of sight.

Naming systems of power. Drawing on Manuel Castells’ network power theory, Actor-Network Theory, and economic sociology of market-making, the paper provides the most complete structural account we’ve found of how AI concentrates power and why surface-level critiques tend to reinforce rather than challenge it. It belongs alongside Alkhatib, Nelson, and tante as load-bearing intellectual infrastructure for our AI ethics work.

POSIWID and decoy accountability. The five decoys show precisely why the POSIWID standard is necessary. Decoys generate the appearance of accountability: safety conferences, regulatory testimony, participatory design sprints. Each appears to offer a tractable path. Each routes critics away from the network of power that makes AI possible. Evaluating by effects — not by claims — is the only exit from the trap.

The safety decoy and disability justice. AI companies yoke “AI Safety” to existential risk while communities bearing the highest cost of AI deployment have the least say in where “safety” gets directed. The paper documents the mechanism with precision. This directly supports our structural critique of AGI doomsday narratives as accountability deferral.

The participatory design decoy and nothing about us without us. This is the most consequential alignment. Vertesi et al. document how stakeholder engagement can be conscripted into inevitability — functioning as ritual that upholds the Project’s march rather than contesting it. This is not an argument against community participation. It is a structural argument for distinguishing genuine power-sharing from consultation theater. Our co-design political demand and nothing about us without us standard are both strengthened by this analysis: they name exactly the form of participation that escapes the decoy.

Where this paper stops. Vertesi et al. don’t write from or about disability, neurodivergence, or access. The political economy analysis operates at a different register than the access and daily functioning analysis. Both are true. Neither cancels the other. A person navigating an inaccessible system with AI assistance is not enrolled in the Project of AI by doing so. The structural critique and the survival imperative coexist. We hold both.

The paper was produced by elite academic institutions. Its citations are primarily elite academic and industry sources. It doesn’t center the communities most harmed by AI. We cite it for the structural mechanism it provides — not as a disability justice document. The disability justice framing is ours to supply.

---

Alignments between Stimpunks AI and Disability Justice guidelines and “Magnifica Vita Mesocosmica: Catholic Ontology in the Era of Artificial Intelligence”

Stefan Ecks. Living Value Theory, livingvaluetheory.org. 2026.
https://livingvaluetheory.org/article/magnifica-vita-mesocosmica-catholic-ontology-ai

Stefan Ecks reads Pope Leo XIV’s encyclical Magnifica Humanitas (May 2026) through the analytical architecture of Living Value Theory (LVT), which proposes that human life is constituted by five irreducible mediations: embodiment, being-with, dwelling, multimateriality, and multisymbolization. The article is written from philosophical ontology and Catholic social thought. Our guidelines are written from inside the neurodivergent and disabled communities these systems most affect. That asymmetry is the starting point. The convergences are substantive.

Transactive dualism is the broken systems argument at the ontological level. LVT names the core pathology of AI capitalism “transactive dualism”: the recursive organization of social life around symbolic-material systems that treat everything else as input. When efficiency becomes the supreme criterion, when people are evaluated by measurable output, when care becomes service delivery, when work becomes productivity — that is transactive dualism operating. Ecks identifies this as the encyclical’s strongest and most accurate diagnosis. The naming is identical in substance to what this page calls the technocratic paradigm, and identical in mechanism to what we call the pathology paradigm in disability contexts. The deficit model that frames neurodivergent people as broken rather than different is transactive dualism applied to minds: normative symbolic and behavioral performance is treated as ontologically primary, and deviation from that standard is coded as disorder, noncompliance, something to correct. LVT supplies the deeper-level argument for why this framing is wrong. Symbolization is not the ground of human worth. It is one emergent mediation among five, dependent on embodiment and dwelling that preceded it evolutionarily by hundreds of millions of years. Broken systems, not broken people — stated as ontological fact.

Recursive suppression, not replacement — the reframing both share. LVT’s most directly useful argument is this: the danger of AI is not ontological replacement. It is recursive suppression. The progressive institutional marginalization of mediations that remain fully operative but are rendered invisible, illegitimate, or uncountable within the dominant symbolic order. Burnout, in LVT’s frame, is not the disappearance of embodiment. It is what happens when embodied rhythms are recursively denied administrative recognition while institutions keep metabolically depending on them. Loneliness is not the disappearance of being-with. It is what happens when symbolic infrastructure organizes being-with around transactional exchange rather than the non-transactional co-presence humans metabolically require. This page makes the same reframing from the community side. The problem AI poses for Autistic and neurodivergent people is not replacement. It is that AI systems encode the recursive suppression of their modes of being, communication, and knowing — and then claim to serve them. Engineered exclusion is recursive suppression operationalized at the pipeline level. The double empathy failure in AI is recursive suppression applied to communication itself: the model’s architecture denies recognition to modes of expression that remain fully present and valid. Neither framework is being dramatic about AI. Both are locating the harm precisely, at the level of institutional design and recognition, where accountability can actually land.

Mediational suppression and the full AI stack. LVT argues that what AI ideology calls “seamless” is achieved by suppressing mediational visibility. The data centers, the rare earth mines, the content moderation workers, the energy infrastructure — these mediations were never absent. What changed was the recursive organization of institutional attention that rendered them invisible as a condition of the system’s symbolic authority. This is the same argument Nelson makes with “the full AI stack,” which this page cites directly. LVT adds the ontological mechanism: you cannot suppress a mediation. You can only deny institutional visibility to one that keeps operating. The harm accumulates in the suppressed register while the symbolic layer claims immateriality. Both frameworks name the structural suppression of these dependencies as intentional, not incidental — the predictable result of a system organized to perform its own immateriality as a condition of its symbolic authority.

POSIWID and mesocosmographic monitoring. LVT’s prescription is ongoing “mesocosmographic monitoring”: diagnosis of which mediations are being suppressed within specific recursive configurations, which recursivity levels are hypertrophying, where symbolic overreach is generating metabolic unsustainability. The encyclical lacks this diagnostic vocabulary. Its prescriptions slide between the specific and the generic because it cannot specify what governance should actually target. This is exactly what Stimpunks’ POSIWID standard does at the community scale. Evaluate tools by what they actually do to the people who need them most — not by product claims, accessibility marketing, or stated commitments. Evaluate by effects on nonspeaking users, AAC users, people whose communication is variable, multimodal, and state-dependent. Both are performing the same diagnostic operation at different scales. LVT’s vocabulary supplies a theoretical grounding for why POSIWID is the right standard: you cannot evaluate by effects what has been rendered institutionally invisible by design, so you must actively restore mediational visibility — which is what evaluating by effects on the most excluded does.

Power concentration, monoculture, and synchronized fragility. LVT reads the Babel metaphor structurally, not theologically: one language, one protocol, one recursive center produces enormous efficiency and enormous fragility. Symbolic monoculture eliminates the redundancy that protects against shock. The efficiency gains of uniformity come with catastrophic synchronized failure potential across biological, economic, and linguistic systems alike. This is the same structural observation Stimpunks makes through Alkhatib’s political definition — AI as an ideological project to shift authority toward centralized structures of power — and through Vertesi et al.’s account of how the Project of AI builds network power while decoys route critics away from it. LVT arrives at the same position from a different angle: concentrated recursive authority is not a contingent feature of bad actors. It is a structural consequence of how transactive dualist systems develop. The people with the most power within the system have the most to lose from its disruption. One practical implication: the case for diversity in communication modes, neurodivergent cognition, and community-led co-design is also a case against synchronized fragility. Monoculture at the cognitive and communicative level is structurally dangerous in exactly the way LVT describes. “Nothing about us without us” is a justice demand. It is also an argument for systemic resilience.

Finitude as condition, not defect. LVT’s argument about finitude is one of its most important convergences with this page. Human limitation is not accidental to the human form of life. It is the condition under which the specifically human form of coordinated adaptation became possible. The metabolic fragility built into embodied dwelling, the vulnerability built into being-with, the finitude of dwelling in a particular place and time — these are not defects in an otherwise perfectible design. They are the evolutionary conditions under which wisdom, compassion, and genuine relation become possible. Any system that treats limitation as a problem to optimize away misunderstands the structure of human development. This is counter-deficit framing stated as evolutionary ontology. Stimpunks makes the same argument from the neurodiversity paradigm: neurodivergent differences are features of diverse minds, not deficits to remediate. The mechanisms converge. A system that treats Autistic communication as deficit does not just fail at representation. It applies transactive dualist logic to human minds — treating symbolic normative performance as the ground of value, and treating deviation as a problem to be corrected. That is the wrong answer. LVT explains why it is structurally wrong, not just politically wrong.

Soul preservation and non-transactive zones. LVT finds genuine value in Christianity’s long institutional maintenance of “non-transactive coordination zones” — ritual, care, the sanctification of ordinary life, the insistence on the dignity of those who produce nothing measurable. These are described as institutional deposits of mediational memory: recursive forms that kept embodiment, being-with, and dwelling visible and institutionally acknowledged within a civilization systematically marginalizing them. Our soul preservation framework is the same operation. The explicit designation of zones that AI does not enter — the glossary, community art, lived experience narratives — is the institutional maintenance of non-transactive coordination. Places where the measure of value is not symbolic output, not productivity, not optimization. LVT would call these the sites where suppressed mediations retain institutional visibility. The art policy holds particular weight here: art from people with actual stakes — with bodies, with sensory histories, with experience of being misread, institutionalized, dismissed — is art that the non-transactive register remains inhabited. An AI has no stakes in those mediations. That is not a minor difference. It is the whole point.

The commons and open licensing. LVT: “The preconditions of coordinated existence are common goods in a deeper sense than economic theory typically acknowledges. They are not goods that belong to anyone because they are the preconditions of anyone belonging to anything.” Stimpunks: since generative AI uses knowledge taken from community, we offer all knowledge and writing derived with AI use to the commons as openly licensed free cultural works. The convergence is exact. The mesocosm cannot be owned because it is the condition of ownership. Knowledge held in common cannot be enclosed without suppressing the mediational ground that made it possible. Open licensing is not just an intellectual property position. In LVT’s terms, it is the appropriate institutional form for knowledge that was always already held in common — a refusal to participate in the recursive suppression of the collective mediational base the knowledge depended on.

Recursivity levels — the vocabulary this page needs. LVT offers something our current AI ethics pages do not yet have by name: a high-resolution framework for recursivity levels. The distinction between first-order effects (specific outputs, individual decisions, particular harms) and second-order effects (the transformation of the conditions under which people form preferences, make judgments, and constitute social reality) is load-bearing for the argument that most current governance frameworks — and most current AI criticism — are operating at the wrong level. AI does not merely influence individual decisions. It reorganizes the recursive architecture within which decisions are made, preferences are formed, and realities are recognized. Most regulation addresses first-order effects. The network of power assembles at the second-order level. LVT’s explicit naming of this distinction gives the sharpest form to what our pages already argue implicitly. The five decoys (Vertesi et al.) are all first-order routers. POSIWID is a second-order corrective. The co-design political demand is a second-order demand. The sycophancy feedback loop is a second-order harm: not one bad answer, but the progressive reorganization of what the user believes they are allowed to think. Engineered exclusion is a second-order harm: not one inaccessible product, but the pipeline-level reorganization of what counts as communication. LVT supplies the unified vocabulary for what all these arguments are pointing at.

Where we diverge.

LVT’s anti-apocalypticism requires a corrective from the community side. Ecks argues the mesocosm will reassert itself regardless — no symbolic system can ultimately abolish embodiment, being-with, or dwelling. The mediations are irreducible. The mesocosm was never at risk of destruction. This is meant to dissolve disproportionate anxiety. It is ontologically correct. And it risks reading as passive. The recursive suppression is causing real harm right now — accumulating across every denied accommodation, every Autistic child handed a compliance-optimization tool instead of a communication scaffold, every nonspeaking person whose expression is structurally inaudible to the system claiming to serve them. The mediations will reassert themselves. People are suffering in the meantime. LVT’s reassurance is ontological. Our urgency is political and material. Both are correct. They do not contradict each other. But the community-accountability corrective is necessary: the mesocosm’s irreducibility is not an argument for waiting. It is an argument for naming what is being suppressed and demanding it be restored to visibility now.

LVT also engages the Catholic tradition on its own terms, with genuine respect for its institutional memory. Our disability justice framework is secular and community-grounded. The theological register is not native here. But LVT’s most important moves — the anti-replacement reframing, the mediational irreducibility argument, the Babel-as-monoculture reading, the recursivity levels framework — are all extractable without the theological scaffolding, exactly as Ecks argues about the encyclical’s own best insights. The mechanism does not require the metaphysics. We take the mechanism.

Finally, LVT operates at the universal ontological scale. Our guidelines operate at the community accountability scale. LVT asks: what are the conditions of coordinated human existence? We ask: who is being suppressed in this specific system, and what do they need to be recognized? Both questions are necessary. The first supplies the theoretical ground for why the second matters. The second supplies the political specificity the first cannot reach from the universal register alone. That asymmetry is not a gap to paper over. It is the relationship: theory provides mechanism, community provides accountability.

Ecks, Stefan. 2026. “The Ontology of Concern: Living Value Theory and the Limits of Catholic Social Doctrine in the Age of Artificial Intelligence.” Living Value Theory. https://livingvaluetheory.org/article/magnifica-vita-mesocosmica-catholic-ontology-ai

---

Alignments between Stimpunks AI and Disability Justice guidelines and “You’ve Been Coded Out: How to Make AI Work for All”

Tetsubayashi is an AI governance and tech ethicist who has spent two decades on trust, safety, and inclusion. She is also a Black West African woman living with sickle cell anemia, born in Togo where most children with disabilities do not reach adulthood. She does not analyze engineered exclusion from the outside. She has been triaged by it. This is the lived-experience version of the structural argument this page makes — the place where testimony and theory meet.

When technology is disconnected from your needs, you’ve been coded out.

Dr. Dédé Tetsubayashi, You’ve Been Coded Out (TEDx)

Coded out is engineered exclusion, named from the receiving end. Tetsubayashi argues that systems fail not because they can’t do the work, but because they’re not designed and built with and for people on the margins. That is Srinivasan’s engineered exclusion stated as testimony. She makes the same move we make: the failure is structural, not personal. It’s a choice product developers make. Not an accident. A choice. Her counter-term — recode us back in — is designed dignity. Exclusion that was engineered in can be engineered out.

Bias is not a glitch. It’s a death sentence. Her central example is an emergency-room triage algorithm built to prioritize the highest-risk patients that instead deprioritized them, because the people who built it had never been sick. This is impact over intent and POSIWID in the same breath. The intent was care. The effect was abandonment. Judge the system by what it does. For people whose access to medical care runs through these systems, the distance between intent and effect is measured in lives.

The most dangerous bias in AI isn’t in the code — it’s in who gets to write it.

Dr. Dédé Tetsubayashi, You’ve Been Coded Out (TEDx)

Accountability does not end at launch. Her second call — work with all possible end users, those on the margins, and keep the communication open after deployment — is our distinction between product-level remediation and pipeline-level redesign. A launch is not a finish line. It is the moment the exclusion becomes measurable. This is “nothing about us without us” as accountability at every stage, not consultation at the testing stage.

Build for one, extend to many. Citing inclusive-design author Kat Holmes, Tetsubayashi argues that designing with the most-excluded person first produces the best outcome for everyone across a product’s entire life cycle. This is the curb-cut logic behind designed dignity: anticipate human variation from the outset rather than retrofit accommodations after exclusion has already happened.

The most-impacted lead. Tetsubayashi is multiply marginalized and a two-decade expert in the field that excluded her. That is the Sins Invalid argument embodied — a disability frame that centers its most-marginalized members, not a narrower thing wearing a broader name.

---

Alignments between Stimpunks AI and Disability Justice guidelines and “AI Hiring Systems Systematically Exclude Neurodivergent Workers”

Paul Hebert, “AI Hiring Systems Systematically Exclude Neurodivergent Workers,” Algorithm Unmasked (June 3, 2025). https://www.algorithmunmasked.com/posts/ai-hiring-systems-systematically-exclude-neurodivergent-workers/

This investigation documents how AI hiring tools — résumé-screening applicant tracking systems, video interview analyzers, gamified personality assessments — filter out Autistic and neurodivergent candidates before a human ever sees the application. It is a civil-rights-framed piece, and a well-sourced one: it draws on Ifeoma Ajunwa’s “The Paradox of Automation as Anti-Bias Intervention,” Barocas and Selbst’s “Big Data’s Disparate Impact,” Sanchez-Monedero et al. in AI & Society, the EPIC complaint against HireVue to the FTC, and the EU AI Act’s classification of recruitment AI as high-risk. Its contribution to our framework is domain: it extends engineered exclusion into employment, the highest-stakes deployment our pages currently reference only in passing.

biased algorithms that mistake neurological differences for incompetence

Paul Hebert, “AI Hiring Systems Systematically Exclude Neurodivergent Workers,” Algorithm Unmasked

Engineered exclusion, moved into hiring. Srinivasan’s frame fits without translation. Résumé screeners penalizing non-linear career paths, video analyzers flagging atypical eye contact, personality tests rewarding neurotypical response clusters — these are the predictable result of choices about data provenance, model objectives, and evaluation practices. Not accidents. Choices. The article names the same structure we name, in a venue we don’t yet cover.

The illusion of objectivity. The article’s vendor critique — that marketing AI as “scientifically objective” obscures the subjective design choices embedded in it — is Ruha Benjamin’s New Jim Code arriving at the hiring funnel. Its own cited source, Ajunwa’s “paradox of automation,” is the legal-scholarship version of the same claim: the tool sold as the cure for bias is the mechanism that launders it. Discrimination dressed as optimization.

Impact over intent. The article notes that the FTC’s enforcement posture can reach discriminatory hiring outcomes even absent intent to discriminate. That is the Algorithmic Justice League’s standard, which we hold directly. Whether a system was built to exclude matters less than whether it excludes.

POSIWID and the algorithmic bottleneck. The article describes candidates eliminated across a sequence — résumé screen, then video analysis, then personality assessment — each stage compounding the last, the rejection invisible and unappealable. That is the purpose of the system being what it does. We evaluate tools by their effects on the people who need them most, not by the objectivity their vendors claim.

Algorithmic double empathy failure. Reduced eye contact read as dishonesty. Communication difference read as inadequacy. The model assumes the candidate is deficient because it was trained on a single, normative profile of what a competent person looks and sounds like. Milton’s double empathy problem, encoded. Neither side is broken. The mismatch is the problem — and the system is the side with the power.

Broken systems, not broken people. The article’s sharpest observation is that candidates are rejected without ever knowing why, with no recourse to contest the decision. That invisibility is exactly the condition under which a person internalizes algorithmic rejection as personal failure. We name the failure structurally so our community doesn’t absorb it as deficit. The rejection is the system’s. Not the candidate’s.

Assess the job, not the interview. The article’s strongest constructive instinct — replace behavioral proxies like “confidence,” “enthusiasm,” and “cultural fit” with objective, task-related performance indicators — aligns with neurodiversity-affirming design. The neurodiversity hiring initiatives it cites at SAP, Microsoft, and JPMorgan, which bypassed traditional interviews their own research showed didn’t work, are evidence for the same thesis that runs through our toolbelt and UDL framing: designing for the margins strengthens the center.

Where more data is not the fix. This is the cleanest divergence. The article’s remedy includes collecting neurological-status data and building “training data that represents the full spectrum of human neurodiversity.” We hold, with Srinivasan, that accessibility is not simply a matter of adding more data — it is a matter of systems capable of reciprocal understanding rather than systems that force users into legibility. More representative data fed into a pipeline built on normative assumptions reproduces the exclusion at higher resolution. The architecture is the problem, not the sample size.

Product-level remediation versus pipeline redesign. HireVue discontinuing facial analysis in 2021 while continuing to analyze speech patterns and word choice is the textbook case our Srinivasan guardrail names: product-level remediation coexisting with infrastructural exclusion. The reform is real. It does not touch the training distributions, evaluation benchmarks, or incentives that produced the exclusion. Removing one biased signal from a system designed around normative legibility is not the same as redesigning the system.

Disability rights, not disability justice. The article works in a civil-rights and accommodation register — the ADA, EEOC technical standards, reasonable accommodation, adaptive interfaces. It asks whether AI can be made to accommodate neurodivergent candidates. We ask the prior question: whether these systems encode the pathology paradigm — the deficit model — before any accommodation is bolted on. It does not reach race, intersectionality, or the neurodiversity paradigm. Accommodation retrofits the system after the exclusion. Designed dignity anticipates human variation from the start.

This is an investigation written for a general and HR audience, not a disability justice document. We cite it for the employment-domain documentation it supplies: the mechanism taxonomy, the HireVue and Amazon cases, the regulatory landscape, the unemployment figures. The structural reading — engineered exclusion, the New Jim Code, impact over intent, the double empathy problem in code — is ours to supply, and it sits one layer beneath where the article stops.

The systematic exclusion of neurodivergent individuals from employment opportunities through biased AI hiring systems represents a civil rights challenge that demands immediate attention.

“AI Hiring Discrimination — Algorithm Unmasked

---

Alignments between Stimpunks AI Collaboration guidelines and “Cognitive Uploading”

Steven Johnson, “Cognitive Uploading” · Michael G. Wagner, “The Epistemology of Cognitive Uploading”

Steven Johnson — co-founder and editorial director of Google’s NotebookLM — names an inversion that runs against the dominant anxiety about AI and learning. The worry is cognitive offloading: hand the machine a generic prompt, take the output, keep the product and lose the education. His counter is cognitive uploading: load a curated corpus of trusted material into a system constrained to reason only over it, and use the tool to provoke judgment rather than replace it. Michael Wagner reads this through Johnson’s “Sleeper Curve” — the method of asking not whether a medium makes us smarter or dumber, but what kind of thinking a given use of it requires. Our AI Collaboration guidelines converge with the architecture here almost exactly, and our Ask page is a working instance of it. On the two questions Johnson leaves open — who owns the tool, and who the method assumes — we go further.

Uploading, not offloading. Johnson’s inversion is our Human-Directed Thinking principle stated from another starting point: AI helps generate possibilities; humans decide what belongs. His sharpest claim is that uploading does not remove friction — it relocates it, away from finding and remembering and toward the harder work of selection and judgment. That is our position on text collaboration. The tool removes a barrier, not the labor. The friction that matters stays with the human. On this we converge completely.

Not an oracle. A guide. Johnson reframes the AI as less a search engine than a connection engine — a place to test relations among sources rather than retrieve a single answer. Our Ask page draws the same line in nearly the same words: not an oracle, a guide. A spider that walks the trails of the garden, never a replacement for it. Wagner’s classroom test — the machine does the retrieval, the student keeps the judgment — is our practice of reading laterally and following the link. The tool’s best answer is an entrance, not a destination.

A corpus you can check. Johnson’s whole case rests on source-grounding: a system bounded to a chosen corpus and made to cite back to passages, so the reader can verify not just whether the output reads well but where each claim is grounded. This is the architecture of our garden-scoped spider. Here we both converge and extend. Johnson’s corpus is private — one mind’s curated library, animated by a model that talks back. Ours is collective: a knowledge garden that is neurodivergent- and disabled-authored, identity-first, and counter-deficit by construction. When the spider answers, the conceptual ground under the answer is ours. Same mechanism; a different politics of whose corpus it is.

Where it stops short: naming the tool. Johnson is building NotebookLM, and the essay is in part product advocacy — Wagner says as much, calling him a builder, not a bystander. Johnson closes on a line we cannot follow: that the difference has never been the machine, it has always been us. We name the systems of power first — AI as a political artifact built to shift authority away from individuals and toward centralized structures. Locating the whole problem in user technique lets the platform off the hook. The uploading/offloading distinction is real and useful, but it is a distinction about how a person uses the wires, not about who owns them. We hold both at once.

Where it stops short: who the method assumes. Johnson and Wagner write about productive struggle, intellectual stamina, friction as the exercise that builds the mind. The unmarked learner underneath that language is one for whom retrieval friction is healthy. Our Ask page is built for a different reader: the seeker who does not yet have the words. You cannot search for monotropism if no one has ever told you monotropism exists. Ground a general-purpose model on the public internet and it returns the pathology paradigm — Autistic accounts of our own lives statistically outweighed by accounts written over us. Epistemic injustice, automated. Scoping the corpus to community-authored knowledge is not about giving a confident reader more to think about. It is about repair. That rationale is absent from both pieces.

The safeguard is structure, not virtue. Johnson trusts the disciplined user to upload rather than offload; Wagner concedes the danger in a closing caution and hopes educators will steer students toward the right path. Our Ask page does not rely on user discipline. The safeguard is architectural: every trail ends at a human-authored page, so every claim the spider makes is one click from the text it claims to summarize. This rhymes with the bounded-competence research our Monotropic AI crosswalk maps — but where that work bounds competence at the level of the model, we bound authority at the level of the interface. The spider carries our voice as far as the gate, then steps aside. Grounding constrains retrieval, not the model’s priors. We say so plainly, on the page.

---

Alignments between Stimpunks AI and Disability Justice guidelines and “Leveraging Health Tech to Revolutionise Accessibility”

Leveraging Health Tech to Revolutionise Accessibility: Advocacy Ally and Digital Passports — Tara O’Donnell-Killen (Thriving Autistic / Luminosity Labs), Autism-Europe, October 2025.

O’Donnell-Killen is an Autistic psychologist presenting a neurodiversity-affirming, community-co-created AI tool: Advocacy Ally, which guides an Autistic person through structured questions about their lived experience and generates concise, rights-based language they can hand to a workplace, clinic, school, or court. Paired with fillable Health and Therapy Passports, the project is one of the closest external convergences with our own practice we have found — human-led, corpus-grounded, epistemic-justice-first, and honest about its limits. It is also a clean illustration of where a well-built affirming tool stops short of the standard we hold, because its core function is translation into the register power already recognizes. We hold it to the same questions we hold our own experiments to.

Autism plus environment equals outcome. The talk’s spine is Luke Beardon’s equation, and O’Donnell-Killen draws the conclusion directly: the difference between thriving and struggling “rarely lies inside the person — it lies in how our systems are designed.” That is our Design Method stated from a podium, and “broken systems, not broken people” carried into health tech.

Co-design as a political demand, not a strategy. The tool was co-created with and for the community; the evaluation was co-led by a neurodivergent team “because epistemic justice matters to us,” with Autistic perspectives shaping every stage of data gathering and interpretation. This is the exact commitment on our disability-justice page — co-design is not a strategy, it’s a political demand — operationalized rather than gestured at.

Corpus-scoped and human-in-the-loop, like the Spider. Advocacy Ally is trained on the community’s own resources and grounded in legal frameworks rather than turned loose on the open web; it retains no data, stores no identifiers, and “supports, not replaces” professional practice. That is structurally the same design as Ask — the Spider in the Garden: a generative tool scoped to a trusted corpus, positioned as a guide rather than an oracle, with a human-authored source always one step away.

Honesty about limits over the demo polish. She flags the Ireland/UK, English-only, internet-dependent sampling; the absence of longitudinal data; and the untested leap from self-reported confidence to material outcomes in employment, health, or education. She calls the tools “prototypes… contributions to a wider conversation,” not solutions. That posture matches our own “these are hard, partly unsolved problems” stance on evaluating relational AI tools.

Where it stops short: translation into the recognized register. O’Donnell-Killen is precise that the tool “doesn’t change what people need — it makes those needs legible,” a bridge from lived experience to “the concise, concrete language that institutions recognize.” That is a real access win, and it is also, structurally, assisted translation into the dominant register. Run against the questions on our relational-tools page: the output is calibrated to what institutions recognize (a double empathy problem the tool addresses from one side), and rehearsing your needs in power’s preferred language sits on the same continuum as masking — adjacent to it, not identical. Asserting an accommodation is not suppressing a behavior; the distinction is real and worth holding. But a tool that trains the flow of adaptation one-directionally, person → institutional legibility, needs to name that direction, or a reader will assume the far bank is already moving.

Where we go further: the far bank is the site of repair. O’Donnell-Killen theorizes the reciprocal move — that systematic advocacy requests could surface recurring needs and drive “organizational learning” and “systemic adaptation” — but files it explicitly as a hypothesis still to be tested. Our ARLES method treats that far bank as the actual work: Attention → Relational/Regulation → Lived Experience → Environment → Systems, present-tense, with the environment and the other party’s obligation to change as the primary demand rather than a downstream maybe. Advocacy Ally builds the bridge and hopes the institution will eventually meet it halfway; we insist the institution is where the redesign has to happen. Same equation, read one layer deeper.

A convergence with an unusually clean edge: a rights-based, non-commercial, Autistic-led tool that does almost everything our guidelines ask, and stops exactly where a general-purpose legibility engine has to stop — at the border its ancestors didn’t theorize crossing.

---

🗺️ Part of our work on AI.