Idam Kim

Ms. Idam Kim

orcid

Research Focus & Metrics

Academic Career Length

2Years (2025-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
30 August 2025 in Education Sciences
Verified
Journal Article
Open access
PDF file
KKKadir Kozan
Kadir Kozan
JHJaesung Hur
Jaesung Hur
IKIdam Kim
Idam Kim
ABAlex Barrett
Alex Barrett
Integrating generative artificial intelligence (GenAI) into professional practice has become an important topic for professional instructional design practice and training. Accordingly, the purpose of this multiple-case study was to examine six professional instructional designers’ integration of GenAI into their professional practice and the factors affecting this integration. Research data were collected through semi-structured interviews conducted with professional instructional designers working in corporate or higher education settings. The results were as follows: (a) instructional designers mostly integrate GenAI into instructional design and/or development phases and they think that it also has the largest impact on these two phases; and (b) instructional designers’ integration of GenAI into their professional practice is mainly based on their ambivalent attitudes toward it, which is closely linked to the advantages and disadvantages associated with the technology. Specifically, instructional designers’ basic understanding of GenAI, the efficiency of generating instructional content through GenAI, the inaccuracy of GenAI-created products, instructional designers’ use of GenAI in everyday life, and institutional or company support shape their attitudes towards and integration of GenAI into their professional practice. All these findings suggest that instructional design and development phases are especially vulnerable to and can benefit from instructional designers’ attitudes and use of GenAI. Accordingly, it can be useful to address and enhance attitudes toward GenAI technology in instructional design training, which can promote instructional designers’ acceptance of the technology and effective use of it.

29 May 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
SHSonghee Han
Songhee Han
JSJueun Shin
Jueun Shin
JHJiyoon Han
Jiyoon Han
BJBung-Woo Jun
Bung-Woo Jun
IKIdam Kim
Idam Kim
ZWZhongyu Wang
Zhongyu Wang
Large language models (LLMs) are increasingly examined as tools for providing feedback during revision tasks, but less is known about when additional structured guidance around such feedback improves task performance. This study examined how users’ baseline justification depth and reflective engagement with feedback from an LLM-based retrieval-augmented (RAG) chatbot shaped performance in a revision task. In a randomized experiment (N = 450), adult participants completed an instructional design task while interacting with the same LLM-RAG chatbot. During the refinement phase, participants received different forms of revision guidance. Participants were assigned to Condition A, no additional scaffold; Condition B, rubric-based scaffold; or Condition C, checklist-based scaffold. Task performance was rated with a three-criterion rubric, and open-ended responses were coded for baseline justification depth and revision theme. Task performance differed across revision conditions: Condition A produced the highest mean performance; Conditions B and C produced lower, nearly identical performance. Baseline justification depth was the strongest predictor of task performance and moderated revision condition effects: predicted performance increased most steeply across all baseline-depth levels in Condition A. However, Condition B showed no evident advantage for participants with lower baseline justification depth. Revision theme also predicted task performance after controlling for baseline justification depth and revision condition. Responses reflecting agency or structured integration of chatbot feedback were associated with higher performance; responses reflecting limited agency or insufficient reflection were associated with lower performance. The findings show that scaffold effectiveness in AI-supported revision depends on users’ baseline justification depth and agentic engagement with AI-generated feedback.