— Karol Jia · scholarly record

Publications
& Talks

Peer-reviewed manuscripts, conference presentations, and works in progress. Most current work sits at the intersection of generative AI, learning technologies, and equity in education. Items marked under review are out for peer review; please request a copy directly.

01 — Manuscripts

Under / Review

2026
First-author manuscript on the emerging educational divide in the age of generative AI.
Currently out for peer review.
under review

An education-side examination of how generative AI is reshaping who gets to learn what, and on what terms. The piece argues that access alone does not close the gap — it can widen it — and looks at what schools and policy can do when a powerful new tool is unevenly distributed and unevenly understood. Title and journal withheld pending review outcome; please email for the manuscript.

02 — Presentations

Conference / Talks

Apr 17, 2026
"When Discipline Becomes Ambiguous Harm: Rethinking Responsibility in School Responses."
School of Education I.D.E.A.S. Conference · Johns Hopkins University · Baltimore, MD.
Feb 28, 2026
"Who Bears the Cost of Verification? Generative AI, Educational Equity, and Policy Challenges Across Educational Systems."
WCCIET 2026 Annual Graduate Student Research Conference · George Washington University · Washington, DC.
03 — Active research

In / Progress

2026—
Multiagent LLM Scaffolding for Simulated Clinical Consultations.
Johns Hopkins School of Education · ongoing research project.

Built and evaluated a multiagent large language model system that distributes patient interaction, tutoring, checklist monitoring, and final feedback across specialized agents to scaffold simulated medical consultations. Preliminary findings: gains concentrated in communication quality — empathic expression and detailed information gathering — rather than diagnostic correctness.

2025—
Hallucination-Aware Learning (HAL) Framework.
Co-authored research project · medical education.

Co-developing an instructional framework that treats generative AI errors not as failures to suppress, but as designed opportunities to strengthen learners' verification skills, evaluative judgment, and critical engagement. Integrates productive failure, cognitive friction, and AI-supported learning design.

— request a copy / collaborate

For drafts,
reprints,
or coffee.