Support answers

It's a web platform for interdisciplinary clinical-reporting (handoff/SBAR) training. You sign in, open a debrief, review the patient's medical records, then run a live spoken encounter with an in-character AI clinician, patient, or family member — either giving or receiving a report. Instructors then grade the session against a rubric and sign it off.

For instructor, student, admin · updated 2026-07-05

Sign in → open a debrief from your list → read the medical records (the first screen) → run the live voice encounter with the avatar → end the session and write your summary (a timed self-report) → the instructor grades it against a rubric and signs it off, after which you see your released score. Instructors can also print/PDF the finished debrief.

For instructor, student · updated 2026-07-05

Open Check your device and connection (linked from the sign-in page and from your invitation; the URL is /check). It reports whether you have a secure (HTTPS) connection, WebGPU for the high-fidelity 3D avatar, the WebGL2 fallback, a working microphone, and — new since 14 August — how fast your network is, which is what decides whether the character's voice arrives smoothly. Every problem it finds comes with what to do about it. It never stops you taking part: a device without WebGPU gets the lightweight avatar, and a slow network still lets you read the replies. The device checks run inside your browser; the speed test times a small block of meaningless data and uploads nothing of yours. The microphone test only runs if you press its button. There is a Copy this result button — send that to your instructor if you need help.

For instructor, student · updated 2026-08-14

No. Every patient, chart, and persona is synthetic training data — never a real patient. AI-generated scenarios are explicitly authored as fictional training material.

For instructor, student, admin · updated 2026-07-05

These answers are in the app at /help — linked as Support answers from the sign-in page, the administrator sign-in page, and the device check. You do not need to be signed in to read them, which is deliberate: the person who most needs them is usually the person who cannot get in. There is a search box at the top. If the problem is the avatar or the sound, run Check your device and connection first — it tests your actual device and tells you what to change.

For instructor, student, admin · updated 2026-08-14

With your email address and the password you chose when you enrolled. Both students and instructors use the same Email and password box at the top of the sign-in page. You never receive or type an access key — the system holds yours on your behalf and it never leaves the server. The collapsed 'Or a student access key' box is only for a handful of accounts issued a key before email enrolment existed.

For student · updated 2026-08-11

The same Email and password box at the top of the sign-in page — students and instructors both sign in there; the system knows your role from your account. Administrators have their own Administrator sign-in page (linked quietly at the bottom of the sign-in page). If your account has MFA enrolled, also enter your current TOTP code in the 'MFA code (if enrolled)' field.

For instructor, admin · updated 2026-08-16

Staff accounts can self-enroll TOTP MFA — enrollment returns an otpauth:// QR URI you scan into an authenticator app (Google Authenticator, Authy, etc.). MFA is then required on your next login. Admins already get a shorter privileged idle timeout on their sessions.

For instructor, admin · updated 2026-07-05

Forgotten password: ask your instructor or an administrator to press Reset password beside your name on the Students page. You get an email, choose a new password, and every other session of yours is signed out. Enrolment link refused: links expire 24 hours after they are sent and stop working once used — ask for a fresh one via Invite by email. If it says an account already exists for your address, you are already enrolled: sign in, or ask for a password reset.

For student · updated 2026-08-11

An enrolled student's email and password always create a student session (as does a legacy access key). For staff, your fine-grained role is bound at login from the central admin identity authority (identity keystone) when enabled; if it's unreachable or disabled, the system falls back to your locally stored role. Admins get the shorter privileged idle timeout.

For instructor, admin · updated 2026-08-11

Press Sign out in the top navigation. It deletes your server-side session, clears your cookie, and returns you to the public home page — not to a login box. On a shared lab machine, sign out before you walk away.

For instructor, student, admin · updated 2026-08-11

Yes. Open the link from your invitation on whichever device is handy — usually the computer your email is on — and the enrolment page shows a QR code beside the password box. Point your tablet's camera at it and it opens the same page there. Opening the page does not use the link up: you set your password once, on whichever device you finish on. The link itself works once and lasts 24 hours. It is worth running Check your device and connection on the device you actually plan to train on, not just the one you enrolled from.

For instructor, student, admin · updated 2026-08-14

Go to Debriefs (/ui/debriefs) and open one, or launch one from Assignments. A not-yet-started debrief takes you to the Medical records review screen first; a debrief already underway drops you straight back into the room.

For instructor, student · updated 2026-07-05

The medical-records screen is the first stop on launch — you review the assigned patient(s)' chart there before talking. Opening a patient's chart marks that patient reviewed. The chart is presented in the medical-records system the instructor assigned (a Helix/Epic-style, Cyrus/Cerner-style, or Meridian/Meditech-style skin over one chart engine). When you're ready, follow the link into the run.

For instructor, student · updated 2026-07-05

It depends on the assignment's direction. In character-gives mode the AI clinician hands off to you and probes with follow-up questions; in the student-gives direction the character asks you for your handoff (situation, background, assessment, recommendation) and follows up. The summary screen's framing flips accordingly. The two-run 'both' (give+receive) direction currently runs as the character-gives flow.

For instructor, student · updated 2026-07-05

Press End session & write summary (in either the 3D room or the 2D room). The session is marked complete and you're taken to a timed self-report: you write, from memory, what you conveyed or understood, and rate your confidence at the start vs now (1–5). The timer logs the minutes you spend organizing after the encounter, mirroring real clinical work. Your summary is then AI-compared to reality and merged into the debrief.

For student, instructor · updated 2026-07-05

Its owner and staff for that tenant. Students can only start assignments that target them — by their student reference, individually or as part of a named group. Cohort targeting was removed in August 2026: it matched on a label the enrolment path never wrote, so a cohort-targeted assignment silently reached nobody. The debrief URL is an unguessable id.

For instructor, student · updated 2026-08-11

Yes — from a debrief's detail page there is a Print / PDF view (/ui/debriefs/{id}/print) that renders a clean, print-friendly copy of the debrief. Use your browser's print-to-PDF to save it.

For instructor, student · updated 2026-07-05

In the 3D room, use the rig's built-in Establish mic → Hold to talk push-to-talk button. In the 2D room, tap 'Tap to speak your response' to start browser speech recognition, speak, then it submits automatically — or expand 'Type instead (last resort)' to type. Voice is the primary path; typing is the fallback.

For student, instructor · updated 2026-07-05

The 3D avatar is the full vrai-faces WebGPU rig that textures the real face and lip-syncs — it needs WebGPU and a portrait image. If your device lacks WebGPU, the room seamlessly falls back to the lightweight 2D avatar: a real blinking, lip-syncing portrait that runs in any browser with no GPU. Both speak the character's lines with synthesized voice.

For student, instructor · updated 2026-07-05

The character speaks audio only on purpose — you must listen, as in a real handoff. The line is kept in the page (visually hidden) for screen readers and appears in the post-debrief transcript, but it is not shown as readable text during the live conversation.

For student, instructor · updated 2026-07-05

Text-to-speech needs an ElevenLabs key configured on the server; without it (or during an outage) the encounter degrades gracefully — the 2D room still shows the line and voice endpoints return no audio (HTTP 204) rather than erroring. Check your device isn't muted and, on the device-check page, that your browser/connection is supported.

For student, instructor · updated 2026-07-05

Those are short 'thinking' filler utterances in the character's own voice. The rig plays one to cover the dead air while the AI reply is synthesized (and to keep audio output awake); if the reply is ready sooner it simply cuts in.

For student, instructor · updated 2026-07-05

The 2D room uses your browser's built-in speech recognition, which some browsers lack. Use the 'Type instead' option below the mic button, or open the session in Chrome or Safari for voice. In the 3D room, push-to-talk is transcribed server-side so it works across Mac/Windows/iPad.

For student, instructor · updated 2026-07-05

Go to Avatars & voices (/ui/avatars, staff only). Every debriefer face is shown as a card with a Replace image control, and there is an Upload a face form below for a character you do not have yet. You choose which face an assignment uses in the Debriefing character section when you build or edit it, which also links straight to this page. Remember the difference: replacing updates every assignment using that character; uploading creates a new one that existing assignments will not pick up.

For instructor, admin · updated 2026-08-14

On Avatars & voices, use the Upload a face form: enter a name, choose a gender presentation, select a PNG/JPG/WebP image under 6 MB, and press Upload face. It becomes a selectable debriefer immediately and the upload is written to the audit log. Note this creates a new character — if you meant to update the picture on a face you already assign, use Replace image on its card instead.

For instructor, admin · updated 2026-08-14

Accepted formats are PNG, JPG/JPEG or WebP, up to 6 MB (it's a face image, not a video). GIF was accepted until 14 August and is not any more — it was stored but neither the 3D rig nor the portrait resolver could read it, so a GIF face was silently invisible. Files outside those types, empty files, or files over 6 MB are rejected; Replace image tells you which rule was broken (the older Upload a face form still just returns you to the page). The file is stored under a generated id — your original filename is never used in the path.

For instructor, admin · updated 2026-08-14

Yes — when you author a case in the Character Generator (Scenario Studio), the AI drafts the patient and supporting cast, and each is registered as an avatar. AI-generated portraits carry an 'AI-generated' disclosure badge in the avatar picker (detected from an .ai.json sidecar or an embedded C2PA 'trainedAlgorithmicMedia' manifest). You can still upload a real image over the top on the Avatars page.

For instructor, admin · updated 2026-07-05

Use Replace image on the face's own card in Avatars & voices. This is the one you almost always want: it swaps the picture but keeps the character — same name, same voice, same id — so every assignment already using that character shows the new image, with nothing to re-pick. There is a link straight to it from the Debriefing-character section when you build or edit an assignment.

Upload a face is different: it creates a new character with a new id, and assignments you built earlier will keep pointing at the old one. Use it for a face you did not have before, not for a better version of a face you already use.

Authored (AI) characters start with no portrait and fall back to a generated placeholder head; give one a real photograph the same way. There is no 'regenerate this exact portrait' button — you supply the image.

For instructor, admin · updated 2026-08-14

An avatar without a real portrait can't drive the 3D WebGPU rig, so the room uses the lightweight generated avatar instead — a distinct SVG face driven by the avatar's gender presentation and a per-id skin/hair tone, so each still reads as its own face. Upload a photo to enable the animated portrait/3D rig.

For instructor, admin · updated 2026-07-05

On Avatars & voices, use Preview a voice: pick one of the ElevenLabs voices, type a sample line, and press Preview. You set the actual voice per assignment when you build/edit it. Each seeded persona has a curated default voice; uploaded faces take a gender-matched default until you assign a voice.

For instructor, admin · updated 2026-07-05

The library ships the real v8 persona set — patients, families, and clinical staff (charge nurse, ED attending, hospitalist, ICU RN, respiratory therapist, pharmacist, and more) — each selectable as a debriefer face, with its real portrait when one exists (e.g. Charge Nurse Kim, persona P-004) and a gender-matched voice.

For instructor, admin · updated 2026-07-05

A rubric is the set of scoring criteria an instructor uses at sign-off. The base rubric has five 1–4 criteria: Completeness, Accuracy, SBAR quality, Prioritization, and Safety, combined into a weighted composite score (normalized 0–1). Manage rubrics under Rubrics (/ui/rubrics).

For instructor, admin · updated 2026-07-05

On Rubrics, create a New blank rubric, or Duplicate/Open a copy of an existing one, then edit the criteria — label, scale anchor, and weight — and save. The scale anchor is the short line shown beside that criterion at sign-off (for example, “1 = major omissions · 4 = totally complete”); leave it blank and the grader sees only the general 1–4 legend. The built-in base rubric is read-only, so editing it opens an editable copy — the base itself never changes.

For instructor, admin · updated 2026-07-05

Open the completed debrief's detail page. If you're staff, you get the Instructor sign-off form: pick the rubric, score each criterion 1–4 (each shows its own scale anchor, and every box is required), and submit. That records the composite score and releases the graded rubric breakdown to the student.

For instructor, admin · updated 2026-07-05

A student sees the graded score and rubric breakdown only after the instructor signs off. Until then the debrief shows as awaiting instructor review; the score is gated so it isn't visible early.

For student, instructor · updated 2026-07-05

Yes. For directions where the character gives the report, an instructor-set report quality dial shapes it: Complete (clean SBAR), Typical gaps / assumptions (a realistic rushed handoff with unflagged gaps), or Contains an error (exactly one plainly-stated factual error the student must catch). A separate experience dial (New / Developing / Experienced) shapes report quality and how sharply the character probes.

For instructor, admin · updated 2026-07-05

Assignments → New assignment. Pick the case, patient character, variant (V1–V5), debriefer avatar and voice, direction, medical-records system, and difficulty dials, then tick the students in the Available students list (or Select all for the cohort). Create it while signed in as yourself: the debriefs an assignment produces go to its creator's Debriefs list, and editing someone else's assignment does not transfer that — the Edit page warns you when an assignment belongs to a colleague. As of 16 Aug 2026 each instructor's Assignments and Debriefs pages show only their own courses (administrators see everything), and an administrator's builder has an extra “Instructor” selector — pick a colleague there to create the assignment on their behalf: it lands on that instructor's lists, their name goes in the student's notice email, and they review the debriefs. That selector is the way to hand a group of students to an instructor; there is no separate student-to-instructor binding — whoever's assignments they run is whose student they are.

For instructor, admin · updated 2026-08-16

Yes, as of 15 Aug 2026 — each targeted student is emailed a notice with the case title, the due date if you set one, which instructor assigned it (the creator — the person who will review it), a sign-in link, and a QR code so a student reading the mail on a laptop can open the sign-in page on a tablet by pointing its camera at the screen. Editing an assignment emails only newly added students, so fixing a typo or moving a due date re-mails nobody. If some notices cannot be sent, the assignment still saves and the Assignments page tells you how many went out.

For instructor, admin, student · updated 2026-08-15

Two different jobs, and it is worth knowing which you are doing. To target work at a class: on New assignment, use Upload a class roster — it digests the file into a student list and pre-fills the group field, and creates no accounts. To create the accounts: Students → 'Enrolling a whole class? Upload a class list instead'. Both accept .xlsx, .docx, .csv, .tsv and .txt (old .xls/.doc are not supported).

For instructor, admin · updated 2026-08-11

On the Assignments list, Launch to verify starts a session from that assignment for you (the instructor) — carrying everything you assigned (EHR system, board, character, direction, dials) so you experience exactly what the student will, without affecting their attempt. Students use the same launch to start their own assigned session.

For instructor, admin · updated 2026-07-05

On the Assignments list, use Export on an assignment to download a JSON scenario bundle (the scenario plus its custom records). You re-load it in another instance via Import scenario. This is a portable scenario transfer, not an LMS grade push.

For instructor, admin · updated 2026-07-05

Go to Import scenario (/ui/scenario-import), choose the exported JSON bundle file, and upload it. It registers the scenario and its custom patient records so it becomes assignable, just like the built-ins.

For instructor, admin · updated 2026-07-05

Students → 'Enrolling a whole class? Upload a class list instead →'. Choose the file and press Read the file. You get a preview showing every row and what will happen to it — Will be invited, or Already has an account — skipped — plus warnings for any row that is not a usable email address. Nothing is created until you press the confirm button. Each student then receives their own enrolment email.

For instructor, admin · updated 2026-08-11

Up to 60 per upload. A larger file is refused whole — nothing is created — and the preview tells you before you press the button, so you can split the list. Upload the smaller files one at a time: each upload is independent, and anyone already invited by an earlier file is simply skipped rather than invited twice.

For instructor, admin · updated 2026-08-12

Almost certainly not. Sending is given a fixed time budget so a slow mail server cannot stall the whole app mid-batch. If your relay is throttling — normal when a burst of messages arrives at once — the remaining students are still invited, but their emails are not sent. Those invites are valid. The page shows their enrolment links: send them yourself, or invite those students again a few minutes later once mail has caught up. This is different from 'could not be emailed', which means the mail server actively rejected the message and is worth investigating.

For instructor, admin · updated 2026-08-12

An email column is required — every student is sent their own link, so there must be an address to send it to. A name column is optional and becomes the student reference instructors see on the debrief list; without one, the address is used instead. You no longer have to make the name match anything by hand: the assignment builder and the unified Assign-work screen both list your students as checkboxes (with Select all), and the group box is filled from the same column the access keys were minted under, so the two cannot disagree. You never need to type a student reference on either screen. Names containing a comma — Chang, Woo — are kept whole. One timing note: a class you have only just invited has no access keys until each student finishes enrolling, so their checkboxes appear (and assignments to them are accepted) after they enrol — invite first, assign second.

For instructor, admin · updated 2026-08-16

No. Any address that already belongs to an account is reported in the preview and skipped, never overwritten. If that student cannot get in, use Reset password rather than re-inviting them.

For instructor, admin · updated 2026-08-11

The audit log's assignment.notified entry names the skipped students by their student reference (a student is skipped when no email address is on file, the send fails, or a very large group runs past the send budget). To try again, open the assignment (Assignments → Edit) and press Resend notice emails — it emails everyone the assignment currently targets, so it also covers a student you have just added or whose address was just fixed. A duplicate notice is harmless; the notice carries no password or link that expires.

For instructor, admin · updated 2026-08-16

Open the Character Generator / Scenario Studio (/ui/scenario-studio). Describe the case premise, setting, acuity/trajectory, and optional learning objectives, then Generate draft. Claude Sonnet drafts a full patient (history, chart), bedside narrative, supporting cast, and treatment path for you to review.

For instructor, admin · updated 2026-07-05

No. The generated draft opens in a Review the draft editor where you can change every field — scenario name, setting, pre-brief, patient details, history, bedside narrative, treatment path, cast, and modules. Nothing is saved or assignable until you press Save; then you build the patient's records and assign it.

For instructor, admin · updated 2026-07-05

Yes — with no Anthropic API key set, generation returns a deterministic template draft built from your inputs so the authoring flow always works. You still review and edit it before saving. With a key, Claude Sonnet 4.6 drafts the richer scenario.

For instructor, admin · updated 2026-07-05

After saving a scenario you land on the patient's records builder. Add each item by generating it with AI (describe it), auto-generating a lab panel, or uploading a file the AI reads. Uploads accept .txt/.md/.csv/.tsv/.json/.pdf and images (PNG/JPG/WebP/GIF), up to 8 MB each. Each record can be shown or hidden and given a purpose.

For instructor, admin · updated 2026-07-05

Facility context (/ui/local-context) holds site-wide standing orders, protocols, formulary, and treatment priorities. These aren't patient records — they're supplied to the character as the site's standing orders it follows and shape how the encounter is judged. Add items as types like standing_order, protocol, medication, treatment_priority, or context; toggle each on/off. Text and file uploads (incl. PDF/images) up to 8 MB are supported.

For instructor, admin · updated 2026-07-05

By invitation, one at a time or a class at a time — Students → Invite a student, or Upload a class list. The student receives an email, follows the link, and chooses their own password. The access key is minted at that moment and held by the server; it is never displayed, emailed, or handed over. This is deliberate: there is no point in the process where a permanent credential is legible on a screen or sitting in an inbox.

For admin, instructor · updated 2026-08-11

Staff can open the Audit log (/ui/audit) to see recent actions — logins, sign-offs, rubric changes, avatar uploads, scenario authoring/import, roster uploads, and debrief turns — each with a timestamp, actor, action, and object. It's append-only and staff-only.

For admin, instructor · updated 2026-07-05

Authentication (logins), debrief turns and completions, instructor sign-offs, rubric create/update/delete, avatar uploads, scenario authoring and imports, and roster uploads are all recorded with the acting user and the affected object.

For admin, instructor · updated 2026-07-05

Admin (privileged) sessions use a shorter idle timeout and absolute lifetime than regular staff sessions as a security measure. Non-admin staff and student sessions use the standard, longer timeouts.

For admin · updated 2026-07-05

The same way as students, as of 15 Aug 2026 — by invitation. Instructors → Add an instructor, pick Instructor or Administrator, and they receive an email with a one-time link and choose their own password. No password is created for you to pass on, displayed, or emailed. Until this change the page printed a temporary password for you to pass on by hand; it no longer does, so there is nothing to copy down, hand over, or leave in a screenshot. If the email cannot be sent, the setup link is shown to you once instead so onboarding is not blocked — that link is an invitation to create a password, not a password.

For admin · updated 2026-08-15

Check Email delivery settings and press Send a test email. Delivery uses your institution's own SMTP relay, so a failure there affects student invites too. If the send failed, the Instructors page shows the setup link once, immediately after you add them — if you have navigated away, press Reset password on their row to send a fresh link. Every attempt, sent or not, is in the audit log as staff.invite.sent / staff.invite.not_sent.

For admin · updated 2026-08-15

24 hours from the moment it is sent, and it stops working once used. If a student misses the window, press Invite by email beside their name for a fresh one — it is one click, which is why the window is short.

For instructor, admin, student · updated 2026-08-11

Yes. Revoke beside a pending invite kills the link immediately. Revoking is audited, and a revoked student can be re-invited afterwards.

For instructor, admin · updated 2026-08-11

On the Instructors page, any account with a live link shows it in its Status column — "Awaiting setup · setup link outstanding" (with the expiry) for a new colleague, or "reset link outstanding" for an existing one — together with a Withdraw link button. Pressing it immediately cancels every outstanding link for that address; the emailed link stops working on the spot, and the withdrawal is recorded in the audit log. The account itself is untouched: press Reset password to send a fresh link, or Archive the row if the address was a mistake. Administrators only.

For admin · updated 2026-08-16

Look under the reply for a ▶ Tap to hear button and press it — the audio is there, your browser simply would not start it on its own. Phone and tablet browsers block sound that begins without you touching the screen first, and because each reply loads a fresh page, the tap you made a moment ago does not carry over. Pressing the button plays it, and you can also open "Can't hear it? Show the words" to read the reply. If you are on a tablet and this happens every time, that is the browser being cautious, not a fault in the session.

For student, instructor · updated 2026-08-23

Press and hold again, and speak again — that is the fix, and losing the attempt costs you nothing. Push-to-talk speech is transcribed by a cloud service, and under heavy class load a transcription can occasionally come back empty; the room waits for your next attempt rather than guessing. If it happens repeatedly: in the 2D room use 'Type instead (last resort)'; in the 3D room check the mic level ring is moving while you hold, and re-run Establish mic if it is not. Your conversation so far is unaffected — nothing is lost by trying again.

For student, instructor · updated 2026-08-20

Uploads that aren't PNG/JPG/JPEG/WebP, are empty, or exceed 6 MB are rejected and you are returned to the Avatars page. Re-export under 6 MB in a supported format. A successful upload appears immediately as a new card and is written to the audit log. If you used Replace image, the page tells you exactly why it was refused.

For instructor, admin · updated 2026-08-14

There are now two reasons, and the room tells you which. Device: the 3D rig needs WebGPU and HTTPS — confirm support on Check your device and connection (/check); iPad needs iPadOS 26+ in Safari, Mac needs Chrome/Edge (or Safari on macOS 26+). Network: since 16 Aug 2026 the room also measures your connection first, and sends you to the lightweight room when it is too slow or laggy for live speech — in that case you see a notice saying so, with a Try the 3D room anyway link. No notice means it was the device. Either way the lightweight animated avatar is used and the encounter, the answers and the grade are unaffected — expected, not an error.

For student, instructor · updated 2026-08-16

Upload a .xlsx, .docx, .csv, .tsv, or .txt file — old .xls/.doc aren't supported (save as .xlsx/.docx and re-upload). Empty files, files with no worksheet/body, or unreadable workbooks are rejected with a message. Fix the file and re-upload; the digest only fills the group field, so nothing is assigned until you submit.

For instructor, admin · updated 2026-07-05

Voice preview needs an ElevenLabs API key on the server. Without it (or during an outage) the preview reports that TTS isn't configured and names the provider tried, instead of erroring. The debrief still works — it just shows the script text until a voice is configured.

For instructor, admin · updated 2026-07-05

The 3D rig's speech connection is kept alive with a periodic keepalive frame and long replies are streamed as small audio chunks, specifically to prevent later turns going silent. If it still happens, reload the debrief room to re-establish the connection; a fresh page reconnects the speech channel.

For student, instructor · updated 2026-07-05

Check three things in order. 1) On the Students page, does the invite show as sent? If the email could not be delivered the page says so and shows the link once — the invite itself still exists. 2) Ask them to check junk/spam; the mail comes from your institution's own relay, not ours. 3) Confirm the address on the Students page is the one they actually read. If any of that fails, press Invite by email to send a fresh link — the old one is replaced.

For instructor, admin · updated 2026-08-11

The character's voice needs a steady 0.5 Mbit/s while it is speaking, and the check warns below about 2 Mbit/s because a shared campus network needs headroom. In order of what actually helps: move closer to the wifi access point (delay is mostly about distance and how many people share the point, not the advertised speed); stop other downloads and video calls on the same connection; try a different network if one is available; use a wired connection if you have one. If it still breaks up mid-encounter, press "Can't hear it? Read it instead" and carry on with the written replies — the encounter, the answers and the grade are unaffected. On a very slow connection the page may make this choice for you before the room loads: you'll see a notice naming the network and a Try the 3D room anyway link. If the speed test could not run at all, that is usually a school or hospital firewall rather than the connection.

For instructor, student, admin · updated 2026-08-16

No. Before the 3D room loads, the page runs its own brief network check — a shorter one than Check your device and connection, using the same speed and delay limits but fewer samples, so the two can occasionally disagree; the full check on /check is the one to trust if you want a considered answer. If the connection is too slow or too laggy for live speech right now, you are placed in the lightweight reading room instead — same conversation, same answers, same grade. If you think the network has recovered, or you have moved to a better connection, press Try the 3D room anyway in the notice at the top of the room; that skips the check and clears the earlier result, so the next time you open the room it measures afresh. The check times a small block of meaningless data and uploads nothing of yours, and if it cannot run — or something on the network intercepts it — the 3D room loads as normal: it advises, it never locks you out. Otherwise the result is remembered for about ten minutes so re-opening the room does not re-run it.

For student, instructor · updated 2026-08-16

A blank section is always recorded as MISSED, with the note "Left blank — nothing was written in this section to compare." That is decided by the system, not by the AI: it applies no matter what the comparison text says, because a section you did not write cannot be credited to you. If you leave the entire summary blank, the overall paragraph says so too. Sections you did write are compared normally, and your instructor sees both your words and the comparison before releasing a score.

For student, instructor, admin · updated 2026-08-16

No. All patients, charts, personas, and AI-generated scenarios are synthetic training data, explicitly authored as fictional and never a real patient. The device-check probes also run locally and upload nothing.

For instructor, student, admin · updated 2026-07-05

AI-generated portraits show an 'AI-generated' disclosure badge in the avatar picker. It's determined from an .ai.json sidecar written when the face is exported, or from an embedded C2PA manifest declaring the image as algorithmically generated media — consistent with an AI-transparency posture.

For instructor, admin, student · updated 2026-07-05

Your microphone is only used when you speak. In the 2D room, speech recognition runs in your browser; in the 3D room, push-to-talk audio is transcribed to text (on-device or server-side) and only the resulting text drives the AI turn. The device-check microphone test runs locally and uploads nothing.

For student, instructor · updated 2026-07-05

Your timed self-report is merged into your debrief, which is visible to you and to staff for your tenant, and it's compared by the AI to the actual encounter for feedback. It becomes part of the record the instructor reviews and signs off.

For student, instructor · updated 2026-07-05

No. Every part of the encounter is mapped to published education frameworks. The chart-review-then-live-handoff structure follows how clinical judgment actually works (NCSBN's NCJMM and Tanner's model); the report is structured against SBAR and I-PASS; the AI debrief follows recognized debriefing science (INACSL standards, PEARLS, GAS, and Rudolph's 'Debriefing with Good Judgment'); and the whole loop is built on experiential-learning and deliberate-practice theory (Kolb, Schon, Ericsson) with the NLN/Jeffries Simulation Theory as the design blueprint. See the 'Evidence Base & Guiding References' section for the full mapping and citations.

For instructor, student, admin · updated 2026-07-05

The rubric is designed to be defensible under modern assessment science. A live spoken handoff is a 'shows how' assessment on Miller's Pyramid — a stronger level of competence than a knowledge quiz. Whether a score interpretation is trustworthy is governed by Messick's unified validity framework and Kane's argument-based approach: you gather evidence that the score supports your intended claim. Rubric-based scoring with explicit level descriptors (Jonsson & Svingby) improves consistency and inter-rater reliability. The system supplies the structure; your program supplies the validity argument for your specific use, which is standard for any performance assessment.

For instructor, admin · updated 2026-07-05

The NCJMM is the framework NCSBN built to measure clinical judgment as a layered cognitive process — recognizing cues, analyzing them, prioritizing hypotheses, taking/communicating action, and evaluating the outcome. It was created to support valid, standardized measurement of clinical judgment on the Next Generation NCLEX. In this system it gives the rubric a defensible scoring skeleton: you can grade whether the learner recognized the salient chart cues, analyzed and prioritized the right hypotheses, and communicated appropriate actions — i.e., score the reasoning process, not just the final answer. Primary reference: Dickison, Haerling & Lasater (2019), J Nurs Educ 58(2), 72-78.

For instructor, student, admin · updated 2026-07-05

SBAR (Situation, Background, Assessment, Recommendation) is a structured verbal-report framework that gives a handoff a predictable four-part skeleton so the receiver knows what's coming and what action is requested. I-PASS (Illness severity, Patient summary, Action list, Situation awareness/contingency, Synthesis by receiver) is an evidence-based handoff bundle associated with roughly a 30% reduction in preventable adverse events in a nine-site study. The system uses SBAR as the default report template and the I-PASS 'synthesis by receiver' read-back maps onto the receive-a-report mode; your rubric can require and score each element for presence, order, and completeness. References: Leonard et al. (2004); Starmer et al. (2014), NEJM 371(19).

For instructor, student, admin · updated 2026-07-05

DASH (Debriefing Assessment for Simulation in Healthcare), developed at the Center for Medical Simulation, is a validated instrument that rates debriefing quality across six elements: establishing and maintaining a safe learning environment, structuring the debrief, provoking engaging discussion, identifying and exploring performance gaps, and helping trainees improve. It reported good reliability (ICC ~0.74; Cronbach alpha ~0.89). Those six elements are the QA yardstick for the AI facilitator: it is designed to hold a safe environment, follow an organized structure, provoke genuine Socratic discussion, surface gaps against the chart/handoff, and close with concrete improvement. Reference: Brett-Fleegler et al. (2012), Simul Healthc 7(5), 288-294.

For instructor, admin · updated 2026-07-05

The AI debrief uses advocacy-inquiry from Rudolph et al.'s 'Debriefing with Good Judgment' — it pairs candid, honest observations with genuine curiosity about why the learner prioritized what they did (their 'frames'), rejecting both harsh judgment and vague 'nonjudgmental' feedback. It follows recognized structures: PEARLS (self-assessment, then focused facilitated discussion, then directive teaching) and GAS (Gather-Analyze-Summarize), inside the psychologically safe, framework-anchored process that the INACSL Healthcare Simulation Standards require. The stance is explicitly 'learning, not gotcha.' References: Rudolph et al. (2006/2007); Eppich & Cheng (2015); INACSL Debriefing Process (2021).

For instructor, student, admin · updated 2026-07-05

For nursing, the system aligns to QSEN (six safety/quality domains as knowledge-skills-attitudes) and the AACN Essentials (2021, a 10-domain competency model including clinical judgment and communication). For interdisciplinary use — RN/LPN/NP plus Y1-2 medical students in the same exercise — it aligns to the IPEC Core Competencies (v3, 2023), whose 'interprofessional communication' domain is exactly what a cross-role handoff develops. TeamSTEPPS (AHRQ/DoD) supplies validated teamwork behaviors (closed-loop communication, check-back, shared mental model) as rubric anchors. These let your rubric use ready-made, citable line items that speak accreditors' language.

For instructor, admin · updated 2026-07-05

The complete, grouped, numbered reference list is in the 'Evidence Base & Guiding References' section of this guide, under 'Guiding references.' Each entry gives authors, year, title, venue, and a DOI or URL where findable, organized by theme (clinical judgment; handoff/communication standards; debriefing science; competency frameworks; experiential learning; assessment/measurement science). A small number of entries carry a [verify] flag on a single bibliographic detail (e.g., a page range or ISBN) that the academic team should confirm before formal publication.

For instructor, student, admin · updated 2026-07-05

Yes — it's designed to double as a research instrument. The platform emits structured rubric scores and debrief transcripts tied to synthetic/charted patients, so datasets carry no protected health information (PHI). Because outcome measures map to validated frameworks (NCJMM/Tanner, SBAR/I-PASS, DASH/PEARLS), you can justify them against published science using Miller/Messick/Kane for your validity argument. IRB note: even though the patient content is synthetic, LEARNER performance data may be human-subjects data — route studies through your institution's IRB, de-identify learner records, obtain consent where required, and document your rubric version and any modifications so results are reproducible and auditable.

For instructor, admin · updated 2026-07-05

The instructor-customizable rubric is the experimental surface. You can add or rename criteria, re-weight the NCJMM reasoning layers or SBAR/I-PASS elements, or hold one condition's rubric fixed while varying scenario difficulty to test a deliberate-practice (Ericsson) or entrustment (EPA / ten Cate) hypothesis — for example, framing 'give a clinical handoff unsupervised' as an entrustment-level judgment across repeated encounters. Keep the level anchors explicit so inter-rater and test-retest reliability stay analyzable, and record the rubric version so the study is reproducible and the validity argument (Messick/Kane) is auditable.

For instructor, admin · updated 2026-07-05

Cite MedSim VRAI Cloud as the software/instrument (product name, version, and access date/URL per your institution's citation style), and separately cite the underlying frameworks you actually used from the 'Guiding references' list — e.g., the NCJMM (Dickison et al., 2019), SBAR (Leonard et al., 2004), and DASH (Brett-Fleegler et al., 2012) if your rubric and debrief relied on them. Do not attribute framework claims to the software; attribute them to the primary sources. Report the exact rubric version and any modifications in your methods so others can reproduce it. If your journal requires it, note that patient content is synthetic and that citations flagged [verify] were confirmed with your academic team.

For instructor, admin · updated 2026-07-05