AI Lab michiganvirtual.org/ai

A practice reference from the Michigan Virtual AI Lab

AI accessibility and learner diversity matrix

Where AI meaningfully reduces access barriers, organized by learner need and by who is doing the work.

29 areas of need · 116 role-specific prompts · 54 cited sources

Michigan Virtual Learning Research Institute · September 2026

Every cell opens a full prompt written for that role and that need, plus two sources in APA 7 format. Read down a column to see what a single role can act on. Read across a row to see how one learner's need shows up differently for the student, the family, the classroom, and the system.

The first section uses Michigan's eligibility categories and rule numbers rather than the federal IDEA names, because that is the language a Michigan IEP team will actually be working from. The second section covers learners with genuine access barriers who are not special education eligible. The third covers the design work that, done once, removes the need for most individual retrofits.

Every prompt uses bracketed placeholders where your own material goes. Nothing identifiable should be pasted into a tool that is not covered by a signed data privacy agreement.

On sources: the 54 cited works are a curated set of verified, authoritative resources, reused across cells wherever they genuinely apply. Citations were not generated per cell, because 116 × 2 unique citations cannot be produced without fabricating some of them.

Six things to settle before any of this reaches a student

These are the failure points that turn an accessibility gain into a compliance problem.

AI output is a draft, not an accommodation of record.

If an IEP specifies a qualified interpreter, live captions do not satisfy it. If it specifies braille instruction, a screen reader does not satisfy it. AI reduces the labor of producing accessible material. It does not change what the plan legally requires.

Auto-generated alt text and captions need a human pass.

Generated alt text describes what is in the picture. It does not know why you assigned the picture. The instructional point is the part a person has to add.

Student data needs an agreement before it goes anywhere.

IEPs, 504 plans, evaluation reports, behavior data, and health information are education records. Confirm a signed data privacy agreement covers the specific tool before any of it is pasted into a prompt. Every prompt here uses bracketed placeholders for that reason.

Never let AI issue a standalone label.

Risk flags, behavior classifications, disability predictions, and language proficiency calls stay with the team. AI can organize the evidence a team reviews. It does not get a vote.

WCAG conformance is the floor, not the goal.

A page can pass every automated check and still be unusable. Test with the people who rely on it, and budget for that testing the same way you budget for the license.

Translation quality varies sharply by language.

Machine translation performs well in high-resource languages and poorly in many of the languages Michigan families actually speak. Verify before relying on it for anything consequential.

Current federal deadline, as of September 2026

The ADA Title II digital accessibility rule still requires WCAG 2.1 Level AA for web content and mobile apps a public entity provides or makes available, including through vendors. On April 20, 2026, the Department of Justice issued an interim final rule extending the compliance dates by one year: April 26, 2027 for public entities serving a total population of 50,000 or more, and April 26, 2028 for smaller entities and special district governments.

The standard, the scope, and the exceptions did not change — only the dates. Existing Title II effective-communication obligations apply now, and private litigation is unaffected by the extension. Districts should confirm with counsel which date applies to them, since the threshold turns on the population of the jurisdiction served.

Read the interim final rule in the Federal Register

Show Columns by Role

Show Rows by Area of Eligibility

Special education eligibility categories13 areas
Populations served outside special education eligibility8 areas
Universal access layers8 areas

Special education eligibility categories

The thirteen categories defined by the Michigan Administrative Rules for Special Education. Federal IDEA uses different names for several of these; the Michigan term and rule number are given.

Area of needStudentParentTeacherAdministrator
Specific Learning DisabilitySLD · R 340.1713Dyslexia, dysgraphia, dyscalculia, processing disorders
Speech and Language ImpairmentSLI · R 340.1710Articulation, fluency, voice, receptive and expressive language
Other Health ImpairmentOHI · R 340.1709aADHD, asthma, epilepsy, diabetes, limited strength or alertness
Autism Spectrum DisorderASD · R 340.1715Social communication, restricted interests, sensory response differences
Emotional ImpairmentEI · R 340.1706Withdrawal, anxiety, depression, sustained behavior barriers to learning
Cognitive ImpairmentCI · R 340.1705Federal IDEA term: intellectual disability
Visual ImpairmentVI · R 340.1708Low vision through blindness; includes braille and large-print users
Deaf or Hard of HearingDHH · R 340.1707Any degree of hearing loss that interferes with learning
Deaf-BlindnessDB · R 340.1717Concurrent hearing and vision loss; intervener-supported
Physical ImpairmentPI · R 340.1709Federal IDEA term: orthopedic impairment; motor access needs
Traumatic Brain InjuryTBI · R 340.1716Acquired injury; needs change across recovery
Severe Multiple ImpairmentSXI · R 340.1714Concurrent intellectual, physical, and functional needs
Early Childhood Developmental DelayECDD · R 340.1711Through age seven, when no other category applies

Populations served outside special education eligibility

Learners with real access barriers who are not covered by MARSE. Several are protected under other federal law; all of them show up in the same classrooms.

Area of needStudentParentTeacherAdministrator
English learners and multilingual studentsTitle IIINot a disability category — do not route through special education
Section 504 only, no IEPSection 504Accommodations without specially designed instruction
Gifted, talented, and twice-exceptionalState and local2e learners hold both giftedness and a disability
Students experiencing homelessness or in foster careMcKinney-Vento · ESSAHigh mobility, interrupted enrollment, device instability
Migrant and highly mobile studentsTitle I Part CRecords arrive late, incomplete, or in another language
Medically fragile, homebound, and chronically absentLocal policyInstruction must reach the student, not the seat
Rural and limited-connectivity learnersDigital equityAccess failure looks like disengagement in the data
Trauma-affected learnersCross-cuttingOverlaps every category above; rarely labeled

Universal access layers

These apply to every piece of content for every learner. Built in once, they remove the need for most individual retrofits. WCAG success criteria are noted where they apply.

Area of needStudentParentTeacherAdministrator
Captions and transcriptsWCAG 1.2.2 · 1.2.4Live and recorded
Alt text and described visualsWCAG 1.1.1 · 1.2.3 · 1.2.5Images, charts, diagrams, video
Reading level and plain languageCross-cuttingAdjusts demand without lowering the standard
Keyboard, switch, and voice operationWCAG 2.1.1 · 2.1.2 · 2.5.1The most common motor-access failure point
Color, contrast, and visual densityWCAG 1.4.3 · 1.4.11 · 1.4.12Low vision, color vision deficiency, visual fatigue
Cognitive load and document structureWCAG 2.4.6 · 3.2 · 3.3Real headings, predictable order, one action per line
Accessible assessmentIDEA · ESSAAccommodations must transfer to the testing platform
Device, bandwidth, and offline accessDigital equityMobile-only households are the norm, not the exception

Prompt


        
      

      

        
Further reading