Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-1648
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-10-17T01:10:07.132219Z
status: published
lens: p_space
sector: health
headline: Concept before capability - a validity-led approach to AI in OSCEs
dek: Objective Structured Clinical Examinations (OSCEs) remain central to the assessment of clinical competence, yet the integration of artificial intelligence (AI) into assessment design, delivery, and scoring raises a fundamental question: what evidence of competence are OSCEs intended to generate? This paper argues that AI should be evaluated through the lens of OSCE validity, with particular attention to how it may strengthen or weaken the interpretive chain linking observed performance to assessment decisions. D…
gain_title: (none)
problem_title: Generative AI used in OSCE materials or scoring may gradually reward observable proxies instead of the clinical capabilities the exams are meant to assess, weakening the validity of high-stakes decisions.
trace_subject: (none)
gain_reading: (none)
gain_evidence: (none)
problem_reading: Generative AI used in OSCE materials or scoring may gradually reward observable proxies instead of the clinical capabilities the exams are meant to assess, weakening the validity of high-stakes decisions.
problem_evidence: reward observable proxies rather than the underlying clinical capabilities
quick_read: The paper examines how artificial intelligence could be used in designing, delivering and scoring Objective Structured Clinical Examinations, which are central to judging clinical competence. It applies Messick's unitary view of validity and Kane's argument-based framework to ask how AI affects the chain of inferences linking observed performance to assessment decisions.

The significance is that OSCEs inform high-stakes decisions, so tools that look plausible but drift away from the intended skills could undermine trust in results. The paper is a conceptual proposal published in 2026; it does not report measured outcomes, so how the framework performs in real exams remains untested.
limitation: The source is a conceptual, argument-based proposal that sets out a framework and principles; it presents no empirical results showing how the framework or AI tools perform in actual OSCEs.
tag: Evidence-backed problem
key_points: The authors argue AI should be judged by its effect on OSCE validity, not by its technical capability. | AI may introduce threats across the scoring, generalisation, extrapolation and implication inferences. | Generative AI is singled out for the risk of construct drift in assessment materials and scoring. | Design support and post-hoc analytics are treated as lower-risk uses when paired with construct-focused review and human oversight. | AI-mediated candidate encounters and automated summative scoring are treated as higher-risk and need stronger validity evidence.
rundown: The authors propose a framework that maps AI applications across the OSCE workflow by how close they sit to candidate consequences and by the evidential burden needed to justify them. This sorts uses into lower-risk and higher-risk categories rather than treating AI as uniformly acceptable or unacceptable.

Their guiding principle, concept before capability, holds that assessment communities must first define what clinical competence means and what evidence should count. Only after that should they decide where AI has a legitimate role in high-stakes assessment.
sources:
- peer_reviewed | Advances in Health Sciences Education | https://doi.org/10.1007/s10459-026-10585-1 | 2026-10-16
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

f7b3dd84f28e3f1c6bb6a7da265b38f2c8ea1d9563bfe8ef534d7955860969e3

previous

0000000000000000000000000000000000000000000000000000000000000000