Evidence
We do not make blanket claims about “objective AI scoring”. Here is what every session records today, and what we are measuring with partner programs.
Every learner conversation is saved so faculty can read exactly what was said.
Each faculty-defined rubric criterion receives its own level and score, not just one overall number.
Feedback for each criterion refers to what the learner actually said or missed in the transcript.
Faculty review every attempt and can rescore a session when needed.
AI-assisted scoring supports repeatable practice. Consequential decisions stay with faculty.
Use AI for repeatable practice; preserve faculty judgment for consequential assessment.
We will publish methods, results and limitations as they become available. Until then, this page lists no results — only what we are measuring.
Validation
Agreement between AI rubric scores and faculty scores, clinical fact consistency, and consistency across repeated runs.
In progressLearning outcomes
Pre/post learner improvement and performance after repeated practice.
In progressTransfer
Where available, performance on faculty-rated or standardized-patient encounters.
In progressFaculty efficiency
Time to author, review and debrief compared with local alternatives.
In progressAccess & reliability
Completion rate, browser and device compatibility, technical failure rate, text fallback.
In progressCase studies
One rigorous implementation story is worth more than many vague testimonials.
In progressResearch
Conference presentations and peer-reviewed studies, with methods, limitations and governance.
In progress