This study explores Knowledge Building as a principle-based innovation at an elementary school and makes a case for a principle- versus procedure-based approach to educational innovation, supported by new knowledge media. Thirty-nine Knowledge Building initiatives, each focused on a curriculum theme and facilitated by nine teachers over eight years, were analyzed using measures of student discourse in a Knowledge Building environment–Knowledge Forum. Results were analyzed from the perspective of student, teacher, and principal engagement to identify conditions for Knowledge Building as a school-wide innovation. Analyses of student discourse showed interactive and complementary contributions to a community knowledge space, conceptual content of growing scope and depth, and collective responsibility for knowledge advancement. Analyses of teacher and principal engagement showed supportive conditions such as shared vision and trust in student competencies to the point of enabling transfer of agency for knowledge advancement to students; ever-deepening understanding of Knowledge Building principles; knowledge emergent through collective responsibility; a coherent systems perspective; teacher professional Knowledge Building communities; and leadership supportive of innovation at all levels. More substantial advances for students were related to years of teachers' experience in this progressive knowledge-advancing enterprise.
Knowledge creation in education is a critical practice for advancing collective knowledge and fostering innovation within a student community. Students play vital roles in identifying gaps and collaborative work to improve community ideas from discourse, but idea quality can be suboptimal, affected by a lack of resources or diversity of ideas. The use of generative Artificial Intelligence and large language models (LLMs) in education has allowed work on idea-centric discussions to advance in ways that were previously unfeasible. However, the use of LLMs requires specific skill sets in prompt engineering, relevant to the in-context technique known as Chain-of-Thought (CoT) for generating and supporting improvable ideas in student discourse. A total of 721 discourse turns consisting of 272 relevant question–answer pairs and 149 threads of student discourse data were collected from 31 students during a two-day student Knowledge Building Design Studio (sKBDS). Student responses were augmented using the CoT approach and the LLM-generated responses were compared with students’ original responses. Findings are illustrated using two threads to show that CoT-augmented inputs for the LLMs can generate responses that support improvable ideas in the context of knowledge creation. This study presents work from authentic student discourse and has implications for research and classroom practice.