SINGAPORE: Personalized comments can lend a hand dental scholars expand the behaviour steering talents had to deal with apprehensive or uncooperative youngsters. Alternatively, reviewing video recordings of appointments with sufferers can position a substantial burden on instructing body of workers. A pilot learn about on the Nationwide College of Singapore (NUS) has discovered that transcripts may supply a quicker solution to overview those interactions and tested the potential for synthetic intelligence (AI) in supporting such tests.
The learn about aimed to check the efficacy of video recordings to transcripts within the analysis of behaviour steering in final-year dental scholars treating paediatric sufferers. It additionally aimed to check the consistency and accuracy of transcript reviews carried out by means of human assessors and massive language fashions (LLMs).
The researchers analysed 40 recorded medical encounters involving dental scholars and sufferers elderly between 4 and 12 years. A human assessor evaluated the transcripts the usage of a paediatric behaviour steering scale in the past used to attain the movies. The researchers tailored the size by means of getting rid of the non-verbal components, and so they additionally got rid of those from the present video-based rankings to permit a right away comparability.
Scoring every transcript took a median of 12 mins, while reviewing the corresponding medical video required a median of 73 mins of viewing time. Regardless of the really extensive distinction in time, the transcript and video tests confirmed just right settlement. The authors concluded that transcript-based analysis introduced a much less time-consuming choice to video evaluation, confirmed just right settlement with video-based overview and may strengthen medical instructing.
The method may permit instructing body of workers to study extra medical encounters and supply scholars with extra widespread particular person comments. Video would however stay vital when educators wish to overview tone of voice, gestures, facial expressions and different non-verbal sides of affected person verbal exchange.
Find out about co-author Dr Shijia Hu, head of orthodontics and paediatric dentistry on the college, mentioned in an NUS press unlock: “Those findings let us beef up coaching for our dental scholars by means of offering extra personalized and insightful comments on their interactions with younger sufferers.”
AI scoring no longer but sufficiently constant
Loose and paid ChatGPT fashions had been caused to judge each transcript thrice in order that their settlement with the human assessor and the consistency in their scoring might be tested Each the unfastened and paid variations of ChatGPT accomplished reasonable settlement with the human assessor. Alternatively, each fashions tended to supply other rankings throughout repeated reviews of the similar transcript.
The unfastened model returned an similar rating in all 3 makes an attempt for most effective seven of the 40 encounters. The paid model did so for most effective 4. The findings point out that the 2 general-purpose LLMs examined weren’t sufficiently constant to function unbiased assessors of scholars’ medical verbal exchange.
The NUS researchers however see doable for an AI-supported digital mentor that would analyse transcripts and lend a hand generate individualised comments. The workforce has gained a grant from the Singapore Ministry of Training’s Tertiary Training Analysis Fund for a follow-up learn about comparing AI-generated comments right through medical dental coaching.
The pilot learn about was once according to a rather small collection of encounters at a unmarried establishment and depended on one human transcript assessor. Additional analysis involving higher and extra various samples, more than one assessors and different LLMs is subsequently had to determine the generalisability of transcript-based scoring and decide whether or not AI-assisted scoring and comments might be dependable and helpful to scholars.
The learn about, titled “Comparing the pediatric habits steering of scholars according to precise medical transcripts scored by means of college and massive language fashions: Pilot comparative learn about”, was once revealed on-line on 12 June 2026 in JMIR Clinical Training.
Subjects:
Tags:
Reside webinar
Mon. 27 July 2026
4:00 am EST (New York)
Federico Zunica
Reside webinar
Fri. 31 July 2026
12:00 pm EST (New York)
Dr. Sergio Florencio
Reside webinar
Mon. 3 August 2026
8:00 pm EST (New York)
Dr. Cameron Shahbazian DMD MBA
Reside webinar
Wed. 5 August 2026
1:00 pm EST (New York)
Dr. Francesco Grande
Reside webinar
Thu. 6 August 2026
5:00 pm EST (New York)
Dr. Daniel Fernando Zabaleta Díaz
Reside webinar
Tue. 11 August 2026
1:00 pm EST (New York)
Dr. Bruce McFarlane Qualified Specialist in Orthodontics Fellow: Royal School of Dentists of Canada Diplomate: American Board of Orthodontics
Reside webinar
Tue. 18 August 2026
7:00 pm EST (New York)
Prof. Dr. Wael Att, Dr. Andrea Laureti ITI Pupil Michigan, Dr. Acela Martinez



