Reference
Accessibility
What is supported, what is not, and what a spoken assessment needs you to decide
RolePlay LMS is assessed against WCAG 2.2 Level AA. The Accessibility Conformance Report (ACR) is version 1.3, dated 26 August 2026, and follows the structure of the VPAT 2.5 WCAG edition.
| Result | Of 55 Level A and AA criteria |
|---|---|
| Supports | 32 |
| Partially supports | 21 |
| Does not support | 0 |
| Not applicable | 2 |
Nothing fails outright. Twenty-one criteria only partly support, and the ones that matter to a teacher are described below. One of those twenty-one moved there in version 1.3 rather than being fixed: the report had assessed the language criteria on the assumption that conversations are in English, and they are not. Ask for the full ACR if you need the evidence for an accessibility review or a procurement process — it names every partial result, what specifically falls short, and what kind of evidence sits behind each finding.
Note
WCAG 2.2, not 2.1
Australian universities usually contract on WCAG 2.1 AA through AS EN 301 549. This product is assessed against 2.2 AA, because the Australian Human Rights Commission’s Guidelines on Equal Access to Digital Goods and Services (April 2025) name 2.2 as the minimum. Conforming to 2.2 satisfies a 2.0 or 2.1 requirement as a superset.
The reading and listening options
Every student can use these, in a panel above Start chat. They do not need to disclose a disability, and the choices are saved in that browser and never sent to us.
| Control | What it does | Who sees it |
|---|---|---|
| Text size | Normal, Large, Largest. Scales the root font size. | Every student |
| Contrast | Normal, High contrast. Darkens ink and strengthens hairlines against a white ground. | Every student |
| Voice speed | Slower, Normal, Faster. Changes how fast the character’s recorded voice plays back. | Every student |
| Microphone | Quiet room, Normal, Noisy room, with a level meter and a Test microphone button. Sets how loud a sound must be before it is sent to a live-voice chatbot. | Every student |
High contrast darkens text and lines on a white background. It does not invert the page or change the whiteboard’s named colours, because inverting breaks the board, and the board is where the teaching is.
Warning
Text size does not reach everything
It scales headings, the student’s brief, the options panel, the navigation bar, form fields and the conversation status line. It does not reach buttons or the small status badges, which are set in fixed sizes. Browser zoom works across the whole interface, so no student is left without a way to enlarge the page. The ACR records this criterion as partially supported and the remainder is scheduled for November 2026.
Keyboard
Hold Enter or Space to use the microphone. Releasing the key stops recording, and the microphone also stops if focus moves away. The star rating on the thank-you screen is chosen with the arrow keys.
The activity navigation bar, the reading options and the main buttons show a visible focus indicator; other controls use the browser’s own. On screens 480 pixels wide or narrower the navigation bar becomes a horizontal scroller, and the browser clips the top and bottom of the focused item’s outline there — the indicator is still visible, but weaker than elsewhere.
What a screen reader gets
The conversation status line is a live status region, so Recording… release to send, Unable to access microphone and Time is up are announced without moving focus. Icons are hidden from assistive technology, every control carries a real text label, and the navigation bar marks the current page in its markup.
There is deliberately one announcing region on the conversation screen. The microphone level meter beside it is hidden from assistive technology and says the same thing in ordinary text, because a second region changing several times a second would make the screen unusable to listen to.
Workflow states are words — Awaiting review, Unfinished, Sent to the gradebook — never a colour on its own. That rule is why the palette carries no green.
Languages other than English
The conversation is not restricted to English. Nothing in the product sets a language: the character answers in whatever language the learner writes or speaks, and will switch part-way through if asked. Speech is transcribed without a language being fixed in advance, and transcripts are stored and shown in the original script. This is not a roadmap item — it already happens, on chatbots built without it in mind.
For an EAL/D learner that is a real adjustment rather than a feature: they can rehearse a difficult conversation in the language they think in before doing it in English. For a languages teacher it is the product working as intended in a subject nobody configured it for.
Warning
What the language support does not cover
- The interface stays in English. Buttons, labels, the options panel and the conversation status line are English whatever language the conversation is in.
- Right-to-left scripts are not supported. Arabic, Hebrew, Farsi and Urdu will lay out left-to-right, which is wrong and hard to read. Do not offer this for those languages without testing it yourself first.
- The page declares itself English. A conversation in another language sits inside a document marked
lang="en", so a screen reader announces it with English pronunciation rules. That is WCAG 3.1.2 Language of Parts, and the ACR records it rather than leaving it implied. - Speech-recognition accuracy across languages is untested by us, exactly as it is across accents within English.
- The character’s voice is not a native speaker’s. The synthesised voices carry an English accent into other languages.
What is still only partly supported
These are the ones a teacher or a reviewer is most likely to meet. The ACR lists all twenty.
| Area | What falls short |
|---|---|
| Text size | Reaches most of the interface but not buttons or status badges. Browser zoom covers them |
| Focus indicator | Clipped top and bottom inside the navigation bar at 480px and under |
| Assessor review panel | Every field is properly labelled, but the panel is not a landmark and carries no heading, so a screen-reader user navigating by heading will not find it. The sentence explaining when a mark is required is visible but not programmatically tied to the field. Neither is a Level AA failure; both are scheduled for October 2026 |
| Turn-taking | Conversation time limits are adjustable and can be switched off. The pause the character waits through before it replies is not individually adjustable |
| Language of a passage | A conversation held in another language sits inside a page declared as English, so a screen reader pronounces it with English rules. See Languages other than English above |
What has been tested, and what has not
The report’s findings come from automated scanning, manual keyboard-only operation, contrast computed against the live palette, reflow measured at 320, 480 and 700 pixels, markup and stylesheet inspection, reading the accessibility tree the browser itself builds, and 49 regression assertions held in the codebase.
Warning
Testing not yet completed
- No testing with users with disability. We say so rather than imply a user panel that did not exist.
- No recorded screen-reader and browser test matrix using NVDA, JAWS or VoiceOver. Reading the browser’s accessibility tree establishes what a screen reader is handed; it does not establish what a person hears, or whether the order and volume of it is usable.
- No speech-recognition accuracy testing across different student groups.
Do not treat automated results as proof that every student can use the product. If your institution requires assistive-technology or user evidence, ask for that testing to be commissioned.
The questions a spoken assessment actually raises
Can a student take part without speaking? Yes. Typing is always available, including in live voice, and produces the same transcript you mark.
Can a Deaf or hard-of-hearing student read the character’s replies? The character’s words appear as text during the conversation. Afterwards the student can download or email the transcript, and teachers can open it from the responses list.
What if a student’s room is noisy, or they cannot speak loudly? Live voice sends only sound that rises clearly above the room, so a chair or a cough does not interrupt the character. A student who is not being heard can choose Noisy room, watch the level meter while they speak, or type instead.
Who decides whether speaking is an inherent requirement? The education provider does, through its own process under the Disability Standards for Education 2005 — amended with effect from 1 August 2026 — in consultation with the student. RolePlay LMS provides the options; it does not decide what adjustment is reasonable. Our responsibility is narrower, and we state it as such: the tool must never be the reason an adjustment is impossible.
What can a teacher adjust? The time limit, including removing it; the conversation limit; and whether students may type. Students change text size, contrast, voice speed and microphone sensitivity themselves, without disclosing anything.
Does an automatically suggested mark go straight to the gradebook? No. A teacher must open the conversation, read the evidence and release the mark, and can change both the mark and the feedback. A conversation that cannot be marked automatically is still held for a person rather than passed or failed by default. See Releasing a mark. Completion-only activities work differently: completion is sent when the student finishes.
Warning
Speech-recognition accuracy
Accuracy varies with accent, first language, speech disability and pitch. Where that output informs a mark, it is an equity issue, and we have not completed accuracy testing across different student groups. A teacher must check the transcript and the evidence before releasing a mark. Human review reduces this risk; it does not remove it.
Two smaller notes
Reduced motion: the interface reduces transitions and hover movement when the browser asks for it. The microphone recording pulse and the loading spinner still move, because they show that something is in progress. The microphone level meter also keeps moving — it is a live measurement, not decoration, and hiding it would remove the only signal that the microphone is working.
Activity frame: the host LMS sets the accessible name for the frame containing RolePlay LMS. Moodle and Canvas both title the activity frame. Check this when you test the activity in your own LMS.