Structured human resource development with AI support is associated with higher expert rated recruitment task quality among education graduates
Abstract
Graduate employability programmes are usually evaluated with self-report measures. We compared a structured human resource development programme supported by generative AI with career guidance matched on topics, duration and facilitator contact. Twelve intact cohorts of graduates seeking education-sector posts in Guangdong, China (N = 186), were allocated as whole units, six per condition, crossed with institution, facilitator and schedule. Blinded experts gave a global rating (0–100) of recruitment-task materials produced on site, under invigilation and without AI, before and after the programme. Ratings were made without a pre-specified rubric, which limits their interpretation. Among 168 participants with post-test data (overall attrition 9.7%), the adjusted difference favoured the AI-supported programme by 5.38 points (95% CI 2.76 to 8.01; d = 0.60) and was stable across eight specifications. Four new self-report scales showed smaller differences (d = 0.12 to 0.29), none significant after Holm correction; each scale could detect only differences of about d = 0.4. The effect on expert ratings exceeded the mean effect on self-report by 0.37 standard deviations (95% CI 0.10 to 0.63). Component analyses were exploratory. The design cannot separate the contribution of AI from structured practice and feedback.
Related
- Pharmacists’ Perspectives on Their University Education: A Comparative Study of Graduates from Northern Sweden
- What Does the Pharmaceutical Industry Look for in Pharmacology Graduates? A Thematic Analysis of Semi-Structured Interviews With Pharmaceutical Industry Employees on Their Perceptions of Pharmacology Graduates.
- Success Among Ethnic Minorities and International Medical Graduates in UK Medical Speciality Recruitment: A Longitudinal Analysis of NHS England Data (2021-2025)
- Junior Doctors' Perceptions of Critical Care Education in Medical School and the Association with Confidence in Core Clinical Skills: A Survey of Graduates in Queensland, Australia.
- Employment satisfaction among public health and preventive medicine graduates and associated factors: a cross-sectional survey at Guangxi Medical University
BibTeX
@article{he2026structured,
title = {{Structured human resource development with AI support is associated with higher expert rated recruitment task quality among education graduates}},
author = {Yuda He and Mazuwin Haja Maideen and Zhiyuan Liang},
journal = {Scientific Reports},
year = {2026},
month = oct,
doi = {10.1038/s41598-026-76662-y},
url = {https://www.nature.com/articles/s41598-026-76662-y},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Structured human resource development with AI support is associated with higher expert rated recruitment task quality among education graduates
AU - Yuda He
AU - Mazuwin Haja Maideen
AU - Zhiyuan Liang
PY - 2026
DA - 2026/10/26
JO - Scientific Reports
DO - 10.1038/s41598-026-76662-y
UR - https://www.nature.com/articles/s41598-026-76662-y
AB - Graduate employability programmes are usually evaluated with self-report measures. We compared a structured human resource development programme supported by generative AI with career guidance matched on topics, duration and facilitator contact. Twelve intact cohorts of graduates seeking education-sector posts in Guangdong, China (N = 186), were allocated as whole units, six per condition, crossed with institution, facilitator and schedule. Blinded experts gave a global rating (0–100) of recruitment-task materials produced on site, under invigilation and without AI, before and after the programme. Ratings were made without a pre-specified rubric, which limits their interpretation. Among 168 participants with post-test data (overall attrition 9.7%), the adjusted difference favoured the AI-supported programme by 5.38 points (95% CI 2.76 to 8.01; d = 0.60) and was stable across eight specifications. Four new self-report scales showed smaller differences (d = 0.12 to 0.29), none significant after Holm correction; each scale could detect only differences of about d = 0.4. The effect on expert ratings exceeded the mean effect on self-report by 0.37 standard deviations (95% CI 0.10 to 0.63). Component analyses were exploratory. The design cannot separate the contribution of AI from structured practice and feedback.
SN - 2045-2322
ER -