cv

Basics

Name Hung-Chun Chang
Label M.S. Student in Psychology, National Taiwan University
Email jonathanchang9@gmail.com
Url https://hungchunchang.github.io
Summary I study whether computational models of learners hold up against real learning data, with a focus on LLM-based simulated students, and I build the tutoring systems and social robots I run those studies in.

Research interests

  • Human learning: how people — scientists included — form concepts, and how the concepts we hold shape what we measure and notice.
  • Learner modeling: simulated students, construct validity, and comparing model and human learning.
  • Human-computer interaction: tutoring systems and social robots for education and older adults, built both as tools and as testbeds for learning research.

Work

  • 2026.06 - 2026.08
    Research Intern
    Learning and Educational Technologies Research Unit, Kyoto University
    Prof. Hiroaki Ogata, Prof. Chia-Yu Hsu; Academic Center for Computing and Media Studies.
    • Initiated the collaboration and secured the placement independently, funded by the International PhD/Master's Students Co-Advising Project.
    • Led a project on the validity of simulated students, accepted at ReLEAF@ICCE 2026 (first author).
  • 2026.02 - present
    Validity of LLMs as Simulated Students
    Master's Thesis Research (in progress), Department of Psychology, National Taiwan University
    Advisors: Prof. Hsiu-Ping Yueh, Prof. Feng-Ming Tsao, and Prof. Hiroaki Ogata (Kyoto University).
    • Recast the validity of an ability-conditioned LLM student simulator as a nomological network of four behavioral evidences: profile–behavior consistency, fidelity of error, boundary of competence, and stateful learning.
    • Tested all four concurrently against real per-tier learner data from the PSLC DataShop Fractions 2012 corpus, using zero-shot prompting and a 2×2 fine-tuning ablation (reasoning × history).
    • Found that fine-tuning with reasoning and history simulated students best, yet the competence paradox persisted and the learning mechanism was not reproduced; this first part was accepted at ReLEAF@ICCE 2026 (first author).
  • 2025.02 - present
    Teaching Assistant, Social Robotics
    Department of Psychology, National Taiwan University
    Prof. Hsiu-Ping Yueh.
    • Led 18+ hours of TA sessions across 4 semesters: literature review, user research, prototyping, and programming for social robotic systems.
  • 2022.11 - present
    Research Assistant
    Social Lab Human-Robot Interaction Group, National Taiwan University
    Prof. Hsiu-Ping Yueh, Prof. Weijane Lin; Department of Psychology.
    • Designed the simulated-student evaluation for an intelligent robot tutor (Enterprise Information Systems, 2026) and the manipulation check for a robot-personality study (International Journal of Human-Computer Interaction, 2026).
    • Sole developer of research systems deployed with real users in a museum, a library, and with older adults.
    • Held 2 workshops on multimodal data analysis and voice-based social robots and 4 on social robot programming; mentored 3 undergraduate projects.

Education

  • 2025.02 - present
    M.S.
    National Taiwan University
    Psychology
    • GPA: 3.87/4.3
    • Applied Linear Regression
    • Experimental Design
    • Neural and Behavioral Modeling
    • Advanced Cognitive Neuroscience
  • 2019.09 - 2025.01
    B.S. in Interdisciplinary Studies of Human-Computer Interaction
    National Taiwan University
    University Interdisciplinary Bachelor's Program
    • Minor: Psychology
    • Thesis: Design and Evaluation of a Conversational Robot for the Elderly Based on Social Penetration Theory
  • 2019.09 - 2025.01
    B.S. in Agriculture
    National Taiwan University
    Department of Forestry and Resource Conservation, College of Bio-Resources and Agriculture

Publications

Projects

  • 2025.09 - 2025.12
    ECRobot, a robot tutor (sole developer)
    An intelligent tutoring system with an iPad app (Swift) for handwriting input, a NAO robot (Qi Python) as voice user interface, and a FastAPI backend. Used in a between-subjects learning study.
    • Instantiated prospect theory as an empathetic feedback algorithm for the robot tutor.
    • Designed the student simulation pipeline and evaluation method, and wrote the first draft of the paper.
  • 2025.02 - present
    Social Robot Design Platform (initiator & sole developer)
    A web-based end-user programming tool that implements a design language for social robot behavior, derived from developing 5 social robotic systems.
    • Refined with four groups of student researchers in a research-through-design process over 6 months.
    • Deployed in a laboratory experiment for over a month.
  • 2025.09 - 2025.12
    TouchHear, an AR assistive learning system (sole developer)
    Real-time fingertip tracking on a depth camera that turns printed images into touch-responsive audio surfaces for visually impaired learners.
    • Replaced the experimenter-in-the-loop step of the original protocol with automatic tracking.
    • Built an authoring interface that lets non-programmers map audio to regions of any printed image.
    • Designed the printed template's ArUco markers to carry both depth-plane calibration and project identity, making each sheet self-describing.
  • 2024.03 - 2025.12
    Companion robot for older adults
    Data analysis and front-end development.
    • Operationalized discourse affectiveness into measurable linguistic indicators and validated the manipulation with NLP-based checks.
    • Developed the front-end app that presents the robot's responses.
    • Sustained conversations averaging 15 minutes per session and 30 minutes per participant.

Awards

Skills

Programming & Tools
Python
Swift
Java
MATLAB
LaTeX
FastAPI
React
TypeScript
Research Methods
Construct Validity
Experimental Design
Representation Probing
User Research
Multimodal Data Analysis

Languages

Mandarin
Native
English
Fluent
Japanese
JLPT N3

Interests

Long-distance running
Half marathon
Road running
Cycling