Technology for a Better World

About Me

Hello, I’m Hajun Kim, an HCI researcher.

My research examines how AI shapes human judgment, agency, and social interaction. I design systems in which AI continually refines its understanding of people’s contexts and adapts its support to how people understand AI and what they expect from it.

This understanding informs how these systems balance AI initiative with human orchestration to improve performance while preserving human agency, tacit knowledge, and critical reflection.

Education

2022 Korean Advance Institute of Science & Technology(KAIST)

M.S in Graduate School of Culture Technology
(specialization in Human Computer Interaction)
Advisor: Juhan Nam & Byungjoo Lee

2020 Handong Global University(HGU)

B.S in Information and Communication Technology Convergence (specialization in Human Computer Interaction)
B.S in User Experience Engineering

Work Experience

2026 NEOWIZ (Game Company)

AI Researcher in AI R&D Center
Key Research Contributions:
- Developed novel pipelines for character agents enabling dynamic collaboration (LLMs-based).
- Developed an image generation tool through iterative studies of designers’ workflows with GenAI.
- Fine-tuned generative models (LLMs, Diffusion) and executed ML optimization.

Selected Research

Steering the Discussion: Exploring Opinion Influence in Groups of Humans and AI Agents

(Under review at CSCW 2027)
Hajun Kim, Dawon Lee

Compared goal-directed and deliberative AI agents in discussions with two humans and two AI agents, alongside a four-human baseline, with AI identities undisclosed during discussion.
Conducted a study with 36 university students in 30-minute discussions, tracking opinions, confidence, multiparty interaction, and identity judgments.
Observed exploratory patterns in opinion strength, confidence, and interactions that final pro/con choices alone could miss.
Introduced the steering shadow to trace AI influence across support for a position, confidence, group responses, and interpretations of who contributed.

IdeAmigos: Comparing User-Orchestrated and System-Orchestrated Multi-Agent Brainstorming

(Under review at CHI 2027)
Hajun Kim, Jeongeon Park

Compared user-orchestrated and system-orchestrated brainstorming with the same four AI agents to examine creative outcomes, agency, intent alignment, psychological ownership, and coordination burden.
Conducted a within-subjects study with 16 students with HCI or UX coursework experience, using two counterbalanced 30-minute brainstorming sessions.
User orchestration received higher average user-experience ratings, without statistically significant differences; creative outcomes varied by task, with broadly similar semantic diversity.
Identified design implications for balancing orchestration control with attention, adapting support to tasks and stages, revisiting ideas, and encouraging critical engagement.

Generative AI as a Mediator in Creator Collaboration: Challenges, Design Opportunities, and Concerns

CSCW 2026 Paper
Hajun Kim*, Jini Kim* (co-first author), Yunjae Choi

Examined how differences in expertise, expressive modalities, and creative priorities complicate common ground between game narrative writers and visual designers.
Conducted a multi-method study with 10 narrative writers and 10 visual designers in South Korea, combining preliminary interviews, creative prompting, and participatory design.
Identified opportunities for GenAI to translate creative intent, maintain shared context, and mediate communication, alongside concerns about reputation, trust, and reduced human communication.
Proposed Dual-Space GenAI: a Personal Studio and Shared Workshop that connect individual creative work with collaboration while preserving control over contextual sharing.

Unlocking Creator-AI Synergy: Challenges, Requirements, and Design Opportunities in AI-Powered Short-Form Video Production

CHI 2024 Paper
Hajun Kim*, Jini Kim* (co-first author)

Explored AI-powered Short-form Video Generators (ASVG) as creator collaboration tools via interviews (N=17) and design workshops (N=12).
Analyzed participation workshops, defining three distinct creator types and their challenges.
Conducted participatory workshops, using four distinct stimuli to help creators envision collaborative AI tools.
Identified design opportunities (e.g., inspiration) tailored to the distinct creator workflows and key creator concerns (e.g., identity dilution) for collaborating with AI.

Image Generation Tool for Game IP Consistency & 1.5-Year Longitudinal Study

Project Lead, NEOWIZ

Developed an image generation tool to support IP-consistent art creation across game studios.
Conducted formative interviews with concept artists from eight in-house studios to analyze their needs.
Built web-based tools integrating customized ComfyUI pipelines and fine-tuned models, and conducted a 1.5-year longitudinal study to analyze usage patterns and propose a next-generation tool.

A Multi-Character Dialogue System for Interactive Game Storytelling

Project Lead, NEOWIZ

Developed an interactive multi-character dialogue system that allows users to create evolving stories through dynamic conversations.
Designed an architecture that integrates each character’s unique persona and long-term memory into the dialogue flow.
Implemented an algorithm that determines the most likely responding agent based on user input, ensuring coherent and context-aware interactions.

Personalizing Game Spaces Based on Individual Visual-Motor Skills

Master Thesis, Paper, Book

Developed a novel difficulty balancing technique that personalizes game design from the start, based on user-skill and mental models.
This technique adjusts game parameters by using a player’s pre-measured visual-motor skill (MTA model) to reach a target error rate.
Conducted a comparative user study across three distinct games (SMG) (personal vs. average design), demonstrating personalization achieved designer-intended performance but yielded unexpected player experience results.