LLMs to Support K-12 Teachers in Culturally Relevant Pedagogy: An AI Literacy Example

Synthesis: Explores LLMs to support K-12 teachers in designing culturally relevant pedagogy. An exploratory pilot with four K-12 teachers found the CulturAIEd tool enhanced teachers' confidence in identifying opportunities for cultural responsiveness in learning activities and in making culturally responsive modifications to existing activities. Addresses equity gaps in AI educational tools by centering culturally relevant content.

Key Findings

• CulturAIEd is an Large Language Models (LLMs)-powered system designed to help teachers contextualize AI literacy activities through Culturally Relevant Pedagogy (CRP) principles, combining student demographic characteristics with rubric-driven guidance in the generative process.
• Teachers struggle to implement CRP due to time, training, and resource gaps; the study asks whether LLMs can lower these barriers (RQ1: influence on teachers' ability to design culturally responsive AI literacy activities; RQ2: perceived strengths and opportunities of such tools).
• In an exploratory pilot with four K-12 teachers, CulturAIEd enhanced teachers' confidence in identifying opportunities for cultural responsiveness in learning activities and in making culturally responsive modifications to existing activities.
• Teachers valued the streamlined integration of student demographic information and immediate actionable feedback, which they associated with high implementation efficiency.
• The study positions this as among the first efforts to empower K-12 educators to design culturally responsive AI literacy activities with low effort using LLMs, situating AI as a partner in CRP.

Study Design & Method

The study combines a design component — building CulturAIEd around CRP/CRT frameworks, including a CRT checklist and demographic customization layered into the LLM's generative process — with an exploratory pilot conducted in preparation for a future mixed-methods study. The pilot with four K-12 teachers examined how the tool influenced their confidence, efficiency, and pedagogical strategies in making learning activities culturally responsive. The authors situate the work against Ladson-Billings' CRP framework (academic success, cultural competence, sociopolitical consciousness) and Geneva Gay's Culturally Responsive Teaching.

What this means for practice

• Faculty developers. Give teachers a structured CRP checklist plus a staged adaptation process — independent adaptation, then checklist-guided revision, then tool-assisted refinement — rather than dropping them straight into a generative tool; that three-phase design is what the pilot used and it maps onto the CRT levels of contributions, additive, transformation, and social action.
• Faculty developers. Configure the generation step to take student demographic context as input, since teachers in the pilot valued the streamlined integration of demographic information and immediate actionable feedback as the source of their efficiency gains.
• Instructors. Treat LLM output as a draft to be revised against what you know about your students: the authors warn that models trained on broad data can reproduce cultural clichés or stereotypes, and that a tool like CulturAIEd cannot replace teachers' contextual understanding.
• Administrators. Fund CRP training and planning time rather than assuming a tool closes the gap, because teachers in this study described limited institutional support for CRP and the time it takes to know students beyond surface traits as the barriers they actually face.

Limitations

• The pilot involved four K-12 teachers (N = 4) and is framed as exploratory; the authors state that the confidence and efficiency improvements should be interpreted as promising trends rather than conclusive evidence of efficacy.
• Each participant took part in a single 90-120 minute semi-structured one-on-one Zoom session (pre-survey, adaptation tasks, post-survey, interviews), so there is no follow-up evidence about whether their classroom practice changed afterward.
• The outcomes are self-reported confidence and perceived efficiency with no control comparison and no measure of student learning; the authors state that larger-scale studies with control comparisons are needed.
• The tool was a demo built on gpt-4o-mini-2024-07-18, and the authors flag stereotype reproduction and the risk of "routinizing" CRP as unresolved risks rather than risks their design eliminated.

Citation

Wang, J., Xiao, R., Hou, X., Li, H., Tseng, Y. J., Stamper, J., & Koedinger, K. (2025). LLMs to Support K-12 Teachers in Culturally Relevant Pedagogy: An AI Literacy Example.