Concept Catalyst: Exploring Scrutable Interfaces to Structure K-12 Teacher Interactions with Generative AI

Synthesis: Mansi et al. (2026) introduce Concept Catalyst, a system designed around 'scrutable interfaces' — interfaces that make AI reasoning visible and editable by users. Working with K-12 teachers, the study shows that when teachers can inspect and modify how a Generative AI tool processes their inputs, they report higher trust, greater sense of control, and better alignment with their pedagogical goals. This directly addresses a critical gap in Edtech Platform design: most teacher-facing AI tools operate as black boxes, undermining the Teaching as a professional decision-maker. The Concept Catalyst approach empowers teachers to become co-designers of AI-assisted lesson planning, not just consumers. The work contributes to AI Literacy by demonstrating that 'scrutability' as a design principle can bridge the gap between AI capability and classroom reality. The study also has implications for K-12 education policy, suggesting that AI tools adopted in schools should meet scrutability standards.

What this means for practice

• Instructors. Map the key ideas of a K-12 engineering design challenge before writing prompts, then hand the AI only the concepts you selected rather than asking for a whole lesson at once. Teachers in the study reported lower cognitive load and more accurate prompts when the interface forced that structuring step first.
• Instructors. Treat every generated scaffold as an editable draft: check the underlying prompt, revise it, regenerate it, or discard it before it reaches students. The teachers prized Concept Catalyst over a chat window precisely because the prompts stayed visible and modifiable.
• Instructors. Keep the pedagogical reasoning in your own hands. The reported value came from making teachers' own structuring of the engineering design process explicit, not from the quality of Generative AI output on its own.
• Instructors. Reuse the same representation when district requirements change, updating the concept map rather than rebuilding prompts from scratch; several teachers adopted the tool for this adaptability.

Limitations

• Ten teachers were recruited through the research team's online contacts and snowball sampling, so the sample reflects existing professional networks rather than a representative range of teachers; five had more than 15 years of classroom experience, and the authors themselves call for broader recruitment across geography, culture, and socioeconomic status.
• The study ran on a Miro mockup with a researcher covertly operating ChatGPT behind the scenes (Wizard-of-Oz), so no participant used a working system and actual output quality was mediated by a fixed prompt template.
• All teachers worked on the same well-known bridge design challenge, chosen to make interactions comparable; the authors note that behavior may differ when teachers write for their own projects.
• No students were interviewed about the scaffolds, leaving their reception of the generated content untested.

Citation

Gennie Mansi, Sunni Newton, Roxanne Moore, Meltem Alemdar, Mark Riedl (2026). Concept Catalyst: Exploring Scrutable Interfaces to Structure K-12 Teacher Interactions with Generative AI.