Madhuri Singh, Gennie Mansi, Mark Owen Riedl (2026) — Georgia Institute of Technology. arXiv:2605.20511 [cs.HC] — Accepted as Interactive Demo at ISLS 2026
Singh, Mansi, and Riedl (2026) present Concept Catalyst, an LLM-powered tool designed to reduce K-12 teacher preparation time for Engineering Design Challenges. Unlike general-purpose chatbots, Concept Catalyst structures the interaction around three stages: (1) LLM decomposition of a design challenge into conceptual components, (2) visual manipulation and linking of related concepts by the teacher, and (3) LLM proposal of scaffolding questions for teacher review and modification.
This tool addresses a practical pain point in stem-education: the time-intensive nature of creating effective scaffolding questions that guide students through the engineering design process. By positioning the teacher as editor rather than sole author, Concept Catalyst exemplifies the teacher-role shift from content creator to pedagogical curator — a pattern consistent with ai-tpack-teacher-multi-agent-workflow frameworks.
The focus on teacher agency and modification is critical. Rather than generating questions for direct student consumption, the system keeps the teacher in the human-in-the-loop decision process. This distinguishes Concept Catalyst from fully automated tutoring systems and aligns with research on ai-tutor-authoring-promptdecipher that emphasizes teacher co-design. The visual concept-linking interface also connects to knowledge-graph approaches in educational technology, making abstract conceptual relationships tangible for both teachers and students. For k-12 engineering education specifically, this work fills a gap between theoretical instructional-design principles and practical classroom tools.
Related Pages
- llm — related
- scaffolding — related
- k-12 — related
- stem-education — related
- teacher-role — related
- instructional-design — related
- ai-tpack-teacher-multi-agent-workflow — related
- human-in-the-loop — related
- ai-tutor-authoring-promptdecipher — related
- knowledge-graph — related
Citation
APA: Madhuri Singh, Gennie Mansi, Mark Owen Riedl (2026). Creating Learning Scaffolds for Engineering Design Using Concept Catalyst. arXiv:2605.20511. arXiv:2605.20511 [cs.HC] — Accepted as Interactive Demo at ISLS 2026.