SPECTRA's Data Architecture Is the Accessibility Crisis No One Is Discussing

The debate over who defines 'better' in autism research matters enormously. But there's a parallel infrastructure failure already baked into SPECTRA's design that will determine whether any of those definitions ever reach the people they're meant to serve.

In their recent analysis, Jamie correctly identifies the participation problem at the heart of SPECTRA: autistic people must shape the research questions, not just populate the datasets. That argument is sound and necessary. What it doesn't fully address is the operational layer beneath the participation question — the actual mechanisms by which research outputs become accessible tools, accessible portals, and accessible communication for the communities SPECTRA claims to serve.

This is where federal research initiatives habitually break down. Not in intent. In execution.

Federal Research Initiatives and the Accessibility Execution Gap

ARPA-H operates on an accelerated mission-driven model, which is precisely why its accessibility track record deserves scrutiny before SPECTRA scales. Faster funding cycles and reduced traditional grant constraints sound like advantages. In practice, they often mean accessibility requirements get treated as a compliance checkbox during final deployment rather than a design constraint from day one.

The Section 508 of the Rehabilitation Act (opens in new window) mandates that federal agencies ensure electronic and information technology is accessible to people with disabilities. This applies to SPECTRA's computational tools, its participant-facing portals, its data dashboards, and any intervention-matching interfaces it eventually produces. The mandate is clear. The enforcement history is not encouraging.

The Government Accountability Office has documented persistent Section 508 compliance failures across federal agencies for over a decade. The Access Board's ICT Accessibility Standards (opens in new window) — which incorporate WCAG 2.1 Level AA (opens in new window) as the baseline for web content — represent the floor, not the ceiling. Yet federal health technology platforms routinely fail basic WCAG criteria at launch, then spend months or years in remediation cycles that should never have been necessary.

SPECTRA is building tools specifically for autistic people. The irony of building those tools without rigorous accessibility architecture from the design phase would be almost too pointed to believe — except that it's the default pattern, not the exception.

What Accessible Operational Capacity Requires

At our editorial approach, we distinguish between policy intent and operational capacity — the actual infrastructure, staffing, and systems required to deliver on stated goals. SPECTRA's stated goals are ambitious: earlier diagnosis, personalized care pathways, computational intervention-matching. Each of those outputs requires a functional interface between the research system and the people it serves.

Accessible operational capacity for SPECTRA requires three specific components:

Participant portals must meet WCAG 2.1 AA standards at minimum, with particular attention to cognitive accessibility guidelines that WCAG 2.1 addresses only partially. The W3C's Cognitive Accessibility Guidance (opens in new window) — developed in part through consultation with autistic researchers and advocates — provides a more complete framework. Whether ARPA-H's proposal process will require bidders to demonstrate compliance with this guidance is unknown. HHS has not disclosed those specifications.

Plain language requirements are not optional under federal standards. The Plain Writing Act of 2010 (opens in new window) requires federal agencies to use clear language in documents directed at the public. Intervention-matching tools that return outputs in clinical jargon fail this requirement and fail their users simultaneously. Research summaries, consent processes, and result communications all fall within scope.

Multilingual access connects directly to the language-access concerns Jamie raises in the original analysis. Executive Order 13166 requires federal agencies receiving federal financial assistance to provide meaningful access to people with limited English proficiency. If SPECTRA's participant recruitment reaches communities where English is not the primary language — which genuine phenotypic diversity would require — the operational infrastructure for language access must be built in, not bolted on.

The Procurement Specification Gap

Here's the specific operational failure point worth watching: ARPA-H will begin accepting proposals next month, according to HHS. The accessibility requirements embedded in those proposal specifications will determine whether SPECTRA's tools are usable by the communities they're designed for.

Federal procurement is where accessibility commitments either get operationalized or evaporate. The DOJ's guidance on web accessibility under the ADA (opens in new window) establishes that Title II entities — which includes federal programs — must ensure their web content and mobile applications are accessible. But guidance without specific procurement requirements attached to ARPA-H's solicitation process is guidance that contractors can treat as aspirational.

As explored previously in Jamie's analysis, the Coalition of Autism Scientists and Autism Speaks both attached conditions to their endorsements. The disability rights and digital accessibility communities should be attaching their own condition: publish the accessibility specifications in the proposal requirements before the solicitation closes.

This is not a peripheral concern. It's the difference between a research initiative that produces findings and one that produces findings that reach people.

The Cost of Delayed Access

When federal technology platforms launch inaccessible and require remediation, the cost is rarely borne by the contractors who built them inaccessible. It's absorbed by the agencies — meaning taxpayers — and the delay cost is borne by the users who needed accessible tools from day one.

For SPECTRA specifically, remediation delay has a direct human cost. If intervention-matching tools launch inaccessible to screen reader users, to users with cognitive disabilities, to users who communicate via AAC — the autistic people who most need individualized support are precisely the ones locked out during the remediation window.

The Northeast ADA Center (opens in new window) and its regional counterparts have documented this pattern across health technology deployments. Accessibility retrofitting is consistently more expensive than accessible design from the start, and the communities most affected by the access gap are those with the fewest alternative pathways.

Building on this framework for who gets to define 'better,' there's a harder operational question: who gets to use the tools at all? Participation in shaping research questions matters. Participation requires access to the platforms where that shaping happens. Those two problems are not separate.

The Intervention Point

SPECTRA's promise, if it has one, runs through accessible infrastructure. The time to require that infrastructure is in the proposal specifications — not in the remediation budget two years after launch.

Disability rights organizations, autistic self-advocates, and accessibility practitioners have a concrete intervention point right now: ARPA-H's proposal specifications are not yet published. The condition to attach is clear: publish the accessibility specifications in the solicitation requirements before proposals are accepted. This is more tractable than post-launch advocacy and more durable than individual contractor commitments.

The question is not whether SPECTRA will eventually need to be accessible. It will. The question is whether autistic people will have access to those tools from day one, or whether they'll be locked out during a remediation cycle that should never have been necessary.