01
Cognitive Scaffolding
Train and evaluate models to reason more like humans.
Measure planning, compositionality, and cognitive bottlenecks, then scaffold LLM behavior for reliability and controllability.
Dongyeop Kang (DK)
Assistant Professor
Shepherd 411
dongyeop@umn.edu
Ph.D. Students and Visitors
Past Members
Human-Centered AI · NLP · Cognitive Systems
I build human-centric language technologies—AI as thinking partners for experts.
My work bridges language and cognition to develop novel algorithms, benchmarks, and interaction frameworks that support experts in complex, real-world cognitive workflows.
Current focus: cognitive scaffolding
I design AI that augments humans: models that learn from how people plan, reason, and create; anticipate cognitive bottlenecks; scaffold difficult tasks; and adapt dynamically to expert strategies. See my research statement (technical, non-technical) or my PhD dissertation. I co-direct Minnesota NLP group.
My current research focuses on three pillars.
02
Thinking Assistant for Experts
Build collaboration workflows for high-stakes domains.
Interactive systems and benchmarks that support scientists, lawyers, journalists, and educators in their cognitive workflows.
03
Societal Alignment
Design NLP systems that reflect plural perspectives.
Diversity modeling, personalization, and fair representation of social values in language technologies.