1
Reviewing the patient brief and clinical content for accuracy
AI-generated scenarios can contain clinically inaccurate, outdated, or internally inconsistent detail. You are responsible for the clinical soundness of everything a student sees.
How to check it
- •Open the simulation in the Sandbox and read the full patient brief, diagnosis, clinical setting, current status and key concern end-to-end.
- •Check that vitals, medications, history and terminology are correct for the condition and the student level.
- •Ask the patient a few probing clinical questions in Sandbox and confirm the answers stay medically plausible and on-topic.
How to fix it
- ✓Edit the scenario narrative and key clinical details in Step 1 (Import & Configure) or Step 2 (Review Content) of the wizard.
- ✓Tighten the system prompt/persona so the patient cannot drift into incorrect clinical claims.
- ✓Re-test in Sandbox until every clinical statement is accurate, then return here and mark the item Yes.
2
Testing the simulation as a student
Publishing without running the full student experience is the most common cause of wasted student credits. What looks right in the editor can behave differently in a live session.
How to check it
- •Use the Sandbox (or generate a test access code) and complete the simulation exactly as a student would, start to finish.
- •Speak or type naturally, trigger the assessment, and view the end-of-session report.
- •Test on the device type your students will use (desktop or tablet).
How to fix it
- ✓Note anything that felt off — audio, pacing, avatar behaviour, scoring — and fix it in the relevant wizard step.
- ✓Repeat a full student run after each change until the whole experience is clean.
3
Setting learning outcomes and rubric for the student level
If outcomes are vague or the rubric does not match them, students are assessed against the wrong criteria and reports become meaningless.
How to check it
- •In Step 3 (Assessment & Rubric), confirm each learning outcome is specific, measurable and written in student-facing language.
- •Check that every rubric criterion maps to an outcome and that the level (e.g. first-year vs final-year) is appropriate.
- •Confirm the number of outcomes is realistic for the time limit.
How to fix it
- ✓Rewrite or reorder outcomes and adjust rubric weightings in Step 3.
- ✓Remove outcomes that cannot realistically be observed in the encounter length.
4
Confirming the assessment report scores and reports accurately
Automated scoring is indicative and can mis-attribute or miss student behaviours. A report that scores incorrectly undermines trust and grading.
How to check it
- •After a test run, open the assessment report and compare the scores against what actually happened in your transcript.
- •Check that each rubric line reflects the student’s real performance and that totals add up correctly.
- •Verify feedback text references the right moments in the conversation.
How to fix it
- ✓If scores are off, refine the rubric wording and outcome definitions in Step 3 so the detectors have clearer targets.
- ✓Re-run and re-score a test session, then use Rescore on an existing session if needed to confirm the fix.
5
Ensuring characters keep their assigned roles
A simulated patient or character can “break role” — answering as the clinician, switching identities, or reversing the scenario — which confuses students and invalidates the encounter.
How to check it
- •In Sandbox, deliberately push the character: ask it to swap roles, ask who it is, and steer off-script.
- •Confirm each character keeps its assigned identity, perspective and role label throughout, across multiple conversation paths.
- •For multi-character sims, confirm the right character responds at the right time.
How to fix it
- ✓Strengthen the persona/system prompt with an explicit role lock (e.g. “You are always the patient; never act as the nurse”).
- ✓Re-test the role-switch attempts until the character holds firm.
6
Checking narration and voice playback
Missing, cut-off, overlapping, or mismatched audio breaks immersion and can make a voice simulation unusable — while still consuming credits.
How to check it
- •Run the simulation with sound on, and listen to every narration and patient voice line start to finish.
- •Confirm the spoken audio matches the on-screen text, plays at a sensible volume, and does not overlap or cut off.
- •Test both with and without headphones.
How to fix it
- ✓If audio and text mismatch, correct the script text in Step 2 so it matches what should be spoken.
- ✓Check the selected voice in the wizard and re-generate/re-test any affected lines.
7
Confirming the simulation runs cleanly start to finish
Freezes, loops, or sessions that never reach a proper end state trap students and burn credits without producing a report.
How to check it
- •Complete at least one full run in Sandbox from launch to the end-of-session report.
- •Test the main decision paths and confirm the session concludes correctly and generates a report every time.
- •Watch for any point where the encounter stalls, repeats, or fails to advance.
How to fix it
- ✓If a path stalls or loops, adjust the scenario flow, events, or time limit in the wizard.
- ✓Re-run the problem path until it completes reliably.
8
Removing placeholder, draft, or test content
Leftover “lorem ipsum”, TODO notes, sample names, or your own test messages look unprofessional and can confuse or mislead students.
How to check it
- •Read through every screen students see — patient brief, clinical cards, outcomes, protocols and reminders.
- •Look for placeholder text, draft notes, test values, or duplicate/sample content.
How to fix it
- ✓Replace or delete any placeholder content in the relevant wizard step.
- ✓Confirm names, dates and details are the intended final values before publishing.
9
Reviewing expected credit usage per student session
Live voice and avatar sessions consume credits per minute. Unreviewed time limits or loops can lead to unexpectedly high consumption across a cohort.
How to check it
- •Check the time limit set for the simulation and multiply by your expected number of students to estimate total usage.
- •Confirm the simulation cannot loop or repeat in a way that inflates session length.
- •Make sure your department/institution is aware of the estimated credit usage.
How to fix it
- ✓Adjust the time limit in the wizard to a realistic value for the task.
- ✓Remove any flow that lets a session run longer than intended, then re-test.
Still stuck after thorough testing? For platform-side technical faults, contact
[email protected]. See also our
Terms & Conditions.
