A study analyzing who uses generative AI more, why, and who achieves better performance as a result has been published in the International Journal of Human–Computer Interaction (IJHCI, SSCI), an international journal in the field of human–computer interaction. The research was conducted by Yunhee Lee, a researcher, and Professor Hyunchul Ahn of the Graduate School of Business IT at Kookmin University, together with researchers at the HJ Institute of Technology and Management (HJITM), with HJITM serving as corresponding author.

The researchers analyzed responses from 245 generative AI users in Korea using structural equation modeling. They examined whether how well generative AI fits one’s work (task fit), how well one knows the work (task knowledge), and confidence in handling AI (AI self-efficacy) lead to use and “generative performance,” and how metacognition — the ability to monitor one’s own thinking and level of understanding — changes these relationships.

The analysis showed that AI self-efficacy (β=0.320), task fit (β=0.252) and task knowledge (β=0.168) all significantly increased generative AI use, and use in turn significantly increased generative performance, the production of new ideas and outputs. The finding that people who know their work well also use AI well shows that AI exerts its power on top of expertise rather than replacing it.

Metacognition made a clear difference on one path. The effect of AI self-efficacy on use was significantly larger in the high-metacognition group (β=0.441) than in the low-metacognition group (β=0.105). This means confidence alone is not enough; only people who can check what they know and do not know translate that confidence into actual use. Differences by metacognition were not significant for the task fit, task knowledge and use–performance paths.

The study shows that to increase the impact of generative AI adoption, organizations and educational institutions should match tasks and tools well and develop both employees’ task knowledge and AI self-efficacy. In particular, it suggests that AI training should include not only prompting skills but also metacognitive training in checking outputs and reflecting on one’s own understanding.

HJITM continues joint research with researchers in Korea and abroad on generative AI competencies and performance.