Video transcripts may streamline paediatric dental training

Dental Tribune International

Tue. 21. July 2026

save

SINGAPORE: Personalised feedback can help dental students develop the behaviour guidance skills needed to treat anxious or uncooperative children. However, reviewing video recordings of appointments with patients can place a considerable burden on teaching staff. A pilot study at the National University of Singapore (NUS) has found that transcripts could provide a faster way to evaluate these interactions and examined the potential of artificial intelligence (AI) in supporting such assessments.

The study aimed to compare the efficacy of video recordings to transcripts in the evaluation of behaviour guidance in final-year dental students treating paediatric patients. It also aimed to compare the consistency and accuracy of transcript evaluations performed by human assessors and large language models (LLMs).

The researchers analysed 40 recorded clinical encounters involving dental students and patients aged between 4 and 12 years. A human assessor evaluated the transcripts using a paediatric behaviour guidance scale previously used to score the videos. The researchers adapted the scale by removing the non-verbal elements, and they also removed these from the existing video-based scores to enable a direct comparison.

Scoring each transcript took an average of 12 minutes, whereas reviewing the corresponding clinical video required an average of 73 minutes of viewing time. Despite the substantial difference in time, the transcript and video assessments showed good agreement. The authors concluded that transcript-based evaluation offered a less time-consuming alternative to video review, showed good agreement with video-based assessment and could support clinical teaching.

The approach could enable teaching staff to review more clinical encounters and provide students with more frequent individual feedback. Video would nevertheless remain important when educators need to evaluate tone of voice, gestures, facial expressions and other non-verbal aspects of patient communication.

Study co-author Dr Shijia Hu, head of orthodontics and paediatric dentistry at the faculty, said in an NUS press release: “These findings allow us to enhance training for our dental students by providing more personalised and insightful feedback on their interactions with young patients.”

AI scoring not yet sufficiently consistent

Free and paid ChatGPT models were prompted to evaluate every transcript three times so that their agreement with the human assessor and the consistency of their scoring could be examined Both the free and paid versions of ChatGPT achieved moderate agreement with the human assessor. However, both models tended to produce different scores across repeated evaluations of the same transcript.

The free version returned an identical score in all three attempts for only seven of the 40 encounters. The paid version did so for only four. The findings indicate that the two general-purpose LLMs tested were not sufficiently consistent to serve as independent assessors of students’ clinical communication.

The NUS researchers nevertheless see potential for an AI-supported virtual mentor that could analyse transcripts and help generate individualised feedback. The team has received a grant from the Singapore Ministry of Education’s Tertiary Education Research Fund for a follow-up study evaluating AI-generated feedback during clinical dental training.

The pilot study was based on a relatively small number of encounters at a single institution and relied on one human transcript assessor. Further research involving larger and more diverse samples, multiple assessors and other LLMs is therefore needed to establish the generalisability of transcript-based scoring and determine whether AI-assisted scoring and feedback could be reliable and useful to students.

The study, titled “Evaluating the pediatric behavior guidance of students based on actual clinical transcripts scored by faculty and large language models: Pilot comparative study”, was published online on 12 June 2026 in JMIR Medical Education.

Topics:

Pediatric Dentistry

Tags:

Artificial intelligence Dental education Dental research Paediatric dentistry Practical training

To post a reply please login or register

New York dental schools receive funding to boost specialised oral care

NEW YORK, US: In April, the state of New York awarded US$25 million (€22 million*) to 30 local hospitals, health initiatives and universities to improve ...

Review highlights gaps in undergraduate implant education in UK and Ireland

BRISTOL, England: Although dental implantology has traditionally been regarded as a postgraduate subject, graduating dentists increasingly need to ...

Which universities are top of the QS dentistry ranking for 2026?

SOLNA, Sweden: Quacquarelli Symonds (QS), a leading global higher education analyst, has recently published its annual QS World University Rankings by ...

The minimally invasive mandate: Are UK dental schools keeping pace?

Despite rapid advancements in digital cariology and biomaterials, a translation gap remains between modern clinical capabilities and resources and the ...

Can skill mix expand preventive dental care for children in England?

LONDON, England: The National Health Service (NHS) dental access crisis in England has brought renewed attention to children’s oral health, as caries ...

High-risk medicines in children’s dentistry raise safety concerns, study finds

PITTSBURGH, US: Prescribing high-risk medicines in children’s dentistry remains uncommon but, carries significant safety implications when it does occur. ...

Education

Live webinar
Wed. 22 July 2026
1:00 pm EST (New York)