Moderate evidence
More
Pharmacy students say AI exam practice is easier to reach but less real than human-led
Study title: Artificial intelligence and human-centred OSCE preparation: A national mixed-methods study of UK pharmacy students' comparative perceptions
- Evidence type
- Survey
- Sample
- n = 240
- Sources
- 1 sourcePeer-reviewed
- Published
- 2 min read
- Truvace record
- TRV-2026-2930cite or check it
Study title: Artificial intelligence and human-centred OSCE preparation: A national mixed-methods study of UK pharmacy students' comparative perceptions
A national study of UK undergraduate pharmacy students compared how they rate AI-led and human-centred preparation for Objective Structured Clinical Examinations (OSCEs). Of 361 students who began an online survey, 240 from 31 institutions completed it, and each rated both approaches on eight domains. Human-centred preparation scored higher in six domains, most strongly for authenticity and communication skill development, while AI-led preparation scored higher for accessibility and psychological safety.
Read the study abstract
Objective Structured Clinical Examinations (OSCEs) are standard in United Kingdom (UK) pharmacy education. Artificial intelligence (AI) tools are increasingly used in preparation, but students' perceptions of them have not been compared directly with those of human-centred methods. To compare UK undergraduate pharmacy students' perceptions of AI-led and human-centred OSCE preparation across eight outcome domains, to examine whether the difference varies by year of study, and to explore how and why students combi…

56Gain side
How the two sides scoreClose call
56Problem side
The gain
In a survey of 240 UK pharmacy students from 31 institutions, respondents rated AI-led OSCE preparation higher than human-centred preparation for accessibility and psychological safety.
The problem
In a survey of 240 UK pharmacy students from 31 institutions, respondents rated human-centred OSCE preparation higher in six of eight domains, most strongly authenticity and communication skill development, and focus groups cited AI's non-verbal deficit and clinical unreliability.
A fixed formula sets both scores from the evidence and its sources. Reader votes don’t move them.
What the scores mean
G is the gain side, out of 100. It goes up when the gain is specific, when it measurably helped people and when the sources behind it are strong.
P is the problem side, out of 100. It goes up when the problem is specific, when it measurably hurt people and when the sources behind it are strong.
In brief
The results suggest students see the two approaches as serving different purposes rather than competing. Complementary use (40.4%) and separate use (38.3%) were about equally common, and between 28.8% and 41.7% rated the two identically within a domain. The survey found no evidence that the gap between the approaches varied by year of study after correction for multiple comparisons.
Three focus groups of 18 students described a non-verbal deficit and clinical unreliability in AI practice. They also described using AI first and human practice afterwards. That sequence emerged mainly from the focus groups and is framed by the authors as a hypothesis needing prospective testing against observed performance.
Main points
- 240 students from 31 UK institutions completed the survey, a 66.5% completion rate among 361 who began it.
- Human-centred preparation was rated higher in six of eight domains; authenticity and communication skill development both had an effect size of r = 0.55.
- AI-led preparation was rated higher for accessibility (r = -0.53) and psychological safety (r = -0.42).
- Complementary use of both methods (40.4%) and separate use (38.3%) were about equally common.
- Three focus groups (n = 18) produced four themes: non-verbal deficit, clinical unreliability, phased pedagogical utility and evaluation divergence.
The full story
The study used a convergent mixed-methods design in which interim survey findings informed the focus groups. Respondents rated both modalities on eight single-item domains drawn from deliberate practice and fidelity frameworks. Paired ratings were compared using Wilcoxon signed-rank tests with Holm correction and bootstrap confidence intervals.
Between 28.8% and 41.7% of students gave the two modalities identical ratings within a domain. No evidence emerged that the difference between modalities varied by year of study once multiplicity correction was applied.
Three focus groups with 18 students were analysed using reflexive thematic analysis and integrated with the survey data in a joint display. Both strands supported complementary use, but the idea of AI scaffolding followed by human practice came mainly from the focus groups.
What this doesn’t fix
The survey used a new, non-validated instrument and convenience and snowball sampling, and the authors say the AI-then-human sequence is a hypothesis needing prospective testing against observed performance.
Sources
1 on file- Currents in Pharmacy Teaching and Learningdoi:10.1016/j.cptl.2026.102841
Talk it over
Students in the survey of 240 rated AI-led OSCE preparation higher for accessibility and psychological safety, and human-centred preparation higher in six of eight domains. How would you weigh those against each other?
Pharmacy students say AI exam practice is easier to reach but less real than human-ledhttps://truvace.com/item/artificial-intelligence-and-human-centred-osce-preparation-a-national-mixed-meth
Students in the survey of 240 rated AI-led OSCE preparation higher for accessibility and psychological safety, and human-centred preparation higher in six of eight domains. How would you weigh those against each other?
Answer in the threadThe survey used a new, non-validated instrument and convenience and snowball sampling. How might that affect what the 240 students from 31 institutions reported about authenticity and communication skill development?
Answer in the threadOf the 361 students who began the survey, 240 completed it. Who might be missing from those who finished, and how could that change the ratings?
Answer in the thread
If you have more time
The focus groups of 18 students described using AI first and human practice afterwards. What would a test against observed OSCE performance need to look like to settle whether that order helps?
Answer in the threadComplementary use (40.4%) and separate use (38.3%) were about equally common. How might a pharmacy school you know of plan its practice sessions around that split?
Answer in the thread
ace
What readers say