Maciej Pankiewicz
University of Pennsylvania, PA·Education Research
Many students find mathematics hard, and digital learning games are a promising way to keep them engaged while they practice. But the help these games offer is usually thin: hints are written in advance, the same for everyone, and cannot respond to what a particular student is struggling with.
- Fields
Education Research
Computer Science
Psychology
- Topics
learning games
educational technology
large language models
- Grants
- 1 known grant, through Aug 2028
What they work on
- Extend the Decimal Point mathematics learning game so that a large language model generates hints personalized to each student's problem-solving history, misconceptions and in-game learning behaviors.
- Test whether dynamically generated AI hints improve learning, engagement and productive learning behaviors compared with traditional static hints or no hints, in classroom studies with roughly 800 fifth- and sixth-grade students.
- Determine when, why and for which students personalized hints are most beneficial, and document the practical challenges of deploying large language models inside educational technology.
- Produce design guidelines for educational game developers, publicly available software and research resources, and a freely available enhanced version of Decimal Point for classroom use.
More about the research
Large language models — AI systems that generate text — could instead produce hints tailored to the individual, yet little is known about whether this actually helps learning or how to do it responsibly in a classroom. This project tests that question directly with elementary and middle school students learning about decimals.
Why it matters
Decimals are a persistent stumbling block in school mathematics. If AI-generated hints prove effective, teachers and students could get individualized help inside a free game, extending high-quality tutoring-style support to classrooms that cannot provide it one-on-one. The design guidelines and released software could also help other developers use language models in learning tools responsibly.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Narges Norouzi
University of California-Berkeley, CA·Education Research
Model students' problem-solving strategies and algorithmic thinking in computer science courses by integrating fine-grained process data to trace competencies across levels of abstraction.
Kayce Mastrup
California State University-Monterey Bay, CA·Education Research
Prototype a bounded large language model homework tool for grades 3–5 that uses structured prompting constraints to guide students' reasoning on science and math problems rather than giving direct answers.
Sarah Hansen
Columbia University, NY·Education Research
Compare webcam-based eye tracking with traditional screen-based eye tracking while students solve chemistry problems. Determine whether low-cost webcams capture viewing behavior well enough for research and feedback.
The school
University of Pennsylvania: 90 labs on PI Pages, $2.2B a year on research, 750 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.