Blinded quasi-experimental AI-versus-student assessment marking protocol¶

Design · All designs
Evidence · no claims cited

Description¶

A pilot evaluation protocol in which ChatGPT-generated submissions, formatted by an independent research assistant to resemble student work, were mixed with deidentified past student assessments and marked by blinded discipline-expert academics against standardized rubrics. As printed, "33 assessments, across the 5 assessment types, were selected for the study: 18 were AI-generated (experimental group), and 15 were past students’ assessments (control group)". A second-marker consensus process applied when a remark differed by more than 10 marks.

Design Implications¶

Context¶

Requirements¶

Blinded markers with postgraduate qualifications and DH or HIM expertise; standardized rubrics; deidentification of student work

Constraints¶

Quasi-experimental design without random assignment; limited samples per category; markers' informal AI-authorship judgments were sometimes inaccurate

Target Learners¶

undergraduate and master's-level digital health and health information management students

Learning Goals¶

evaluating assessment validity under generative AI
academic integrity benchmarking

Claims¶

Examples¶

-

Key Sources¶

Wani Tafheem Ahmad, Liem Michael, Prasad Natasha, Robinson Kerin, Nexhip Abbey, Tassos Melanie, Gjorgioski Stephanie, Khan Urooj Raza, Boyd James, Riley Merilyn. (2026). Susceptibility of Assessment Types to AI-Generated Content in Digital Health and Health Information Management Education: Quasi-Experimental Pilot Study. JMIR Medical Education. https://doi.org/10.2196/82988

Page metadata

Field	Value
`type`	`design`
`id`	`blinded-ai-student-marking-protocol`
`title`	`Blinded quasi-experimental AI-versus-student assessment marking protocol`
`description`	`A pilot evaluation protocol in which ChatGPT-generated submissions, formatted by an independent research assistant to resemble student work, were mixed with deidentified past student assessments and marked by blinded...`
`status`	`review`
`generated`	by `process:settle-candidates` on `2026-11-05`
`trust`	unverified — no one has confirmed this page's content
`sources`	1 — `wani-tafheem-ahmad-2026`

Frontmatter is the page's machine-readable record — see the Schema & Guide.