🎙 AI Scribe Missed/Fabricated Detail Detection Simulator
A simulator for detecting missing or fabricated details in notes generated by an AI scribe.
The Conversation Is the Ground Truth
Placeholder lead: the visit conversation is the only true record.
- 16: Details discussed (Placeholder caption text here)
- Audio transcript: Source type (Placeholder caption text here)
- Ground truth: Review basis (Placeholder caption text here)
- Chart accuracy: Downstream risk (Placeholder caption text here)
Why the conversation anchors the audit
Placeholder section body: every later check compares back to this reference list.
The AI Scribe Produces a Draft
Placeholder lead: the model summarizes the conversation into a note.
- AI scribe model: Draft source (Placeholder caption text here)
- Variable: Note details (base) (Placeholder caption text here)
- Physician: Editable by (Placeholder caption text here)
- Clinician of record: Signed by (Placeholder caption text here)
Drafting is not transcription
Placeholder section body: summarization can drop or invent clinical details.
Finding What the Note Left Out
Placeholder lead: omissions hide relevant details behind a clean note.
- 0–30%: Omission rate range (Placeholder caption text here)
- Missing information: Risk type (Placeholder caption text here)
- Side-by-side compare: Detection method (Placeholder caption text here)
- Grey, dimmed: Marker (Placeholder caption text here)
Omissions look like a clean note
Placeholder section body: nothing on the page signals what is missing.
Finding What the Note Invented
Placeholder lead: fabrications add detail that was never actually said.
- 0–15%: Fabrication rate range (Placeholder caption text here)
- Hallucinated content: Risk type (Placeholder caption text here)
- Side-by-side compare: Detection method (Placeholder caption text here)
- Red flag: Marker (Placeholder caption text here)
Fabrications look authoritative
Placeholder section body: invented details read exactly like real ones.
Catching Both Error Types Before Signing
Placeholder lead: review must clear omissions and fabrications together.
- 2 types: Errors to catch (Placeholder caption text here)
- Missed care detail: Omission risk (Placeholder caption text here)
- Billing/legal exposure: Fabrication risk (Placeholder caption text here)
- Physician sign-off: Final gate (Placeholder caption text here)
Sign-off is the last safety check
Placeholder section body: unresolved errors carry real clinical consequences.
Placeholder highlight: careful review protects patients and the record.
A simulator for detecting missing or fabricated details in notes generated by an AI scribe.
2D · HTML5 Canvas 2D · 60 FPS target · runs fully client-side, no install