01
AI in Education
Generative AI for automatic item generation (AIG) and automatic machine translation (AMT), aimed at assessments that develop critical thinking.
Sogang University · Seoul, Korea
MindScale Lab
Measuring Human Minds and Artificial Minds
We study how artificial intelligence can make educational measurement fairer, faster, and more informative — from automatic item generation and automated scoring to adaptive test design and the psychometrics behind international assessments such as PISA.
Hyo Jeong Shin, Ph.D. · Lab Director
40+
Publications
4
PISA cycles
3
Patents filed
6
Researchers
About the lab
Psychometrics for an AI era
MindScale is the educational measurement and data science lab of Dr. Hyo Jeong Shin at Sogang University. The lab's work sits where measurement theory meets machine learning: we build and evaluate methods that let large-scale assessments say more, with less burden on the people taking them.
Before joining Sogang in 2022, Dr. Shin spent seven years at Educational Testing Service, where she co-led the psychometric operations of PISA 2018 and PISA 2022 — work that introduced multistage adaptive testing to PISA Reading and Mathematics and operationalized machine-supported coding of constructed responses across dozens of languages. In 2026 she became the first Korean researcher appointed to the OECD's PISA Technical Advisory Group, and is currently its only member from Asia.
Today the lab brings that operational experience back to questions that matter for Korean classrooms and for international assessment alike: How can generative AI produce items that are not just fluent but psychometrically sound? How should we score short written answers in a language other than the one a model was trained on? What does an adaptive test owe to students whose achievement distribution looks nothing like the calibration sample?
We work with graduate students, research fellows, and undergraduates across Korea and abroad, and we welcome collaborations with schools, testing organizations, and researchers who share these questions.
Research areas
Four lines of work
Distinct questions, one shared toolkit: item response theory, latent variable models, and modern machine learning.
02
Big Data Analytics
Large-scale assessment data — PISA, PIAAC, NAEP, TIMSS — with advanced imputation and machine learning applied to process and response data.
03
Automated Scoring
Supervised learning and classification of natural language responses across languages, together with models of human rater effects.
04
Adaptive Testing
Multistage adaptive designs for heterogeneous student populations, and the psychometric methods that keep their scores comparable.
Where we're heading
AI Measurement Science
Educational measurement is often mistaken for the business of assigning test scores. The harder and more interesting problem is measuring what cannot be read off a scale: the multidimensional competencies — communication, collaboration, creativity, critical thinking — that no single questionnaire or exam captures well.
The lab's longer-term aim is to build out AI Measurement Science as a field: measuring human capability from the process data people leave behind as they work, measuring how well humans and AI systems collaborate, and turning a century of psychometric method on the language models themselves. Evaluating an AI's capability turns out to be, repeatedly, an exercise in understanding human capability more precisely.
News
Recent news
Registration Chair for Festival of Learning 2026 in Seoul
AIED, EDM, and ACM Learning @ Scale met as a single federated conference for the first time in twenty years, hosted in Seoul from 27 June to 3 July. Dr. Shin served as Registration Chair, merging three separate registration…
Sogang Gazette interview on AI, education, and measurement
Dr. Shin discusses PISA, student-led research, AI Behavioral Studies, and her vision for AI Measurement Science in an interview with Sogang Gazette.
Undergraduate capstone project accepted to the AIED 2026 main track
A project that began in the undergraduate course Understanding Big Data and Its Educational Applications (Capstone Design) was accepted as a main-track paper at AIED 2026 — a venue where main-track acceptance is…
Appointed to the OECD PISA Technical Advisory Group
Dr. Shin is the first Korean researcher appointed to the Technical Advisory Group for PISA, the OECD's international student assessment, and is currently its only member from Asia. The group of roughly eight to ten researchers…
Selected work
Recent and representative
An extended two-parameter logistic item response model to handle continuous responses and sparse polytomous responses
Psychometrika, 90(5), 1705–1731
I’m not stuck – I’m learning: Operationalizing self-efficacy trajectories in large-scale AI learning environments
Artificial Intelligence in Education (AIED 2026), LNCS, 185–199
DOIUndergraduate capstone project
STAIR-AIG: Optimizing the automated item generation process through human–AI collaboration for critical thinking assessment
Proceedings of the 20th Workshop on Innovative Use of NLP for Building Educational Applications (BEA 2025)
Developments in psychometric population models for technology-based large-scale assessments
Journal of Educational and Behavioral Statistics, 44(6), 671–705
Interested in joining MindScale?
We look for students who are curious about measurement, comfortable with data, and willing to learn the statistics that make educational claims trustworthy. Prior psychometrics coursework is welcome but not required — undergraduates from the lab's capstone course have gone on to publish at AIED.
