DEVELOPING A VIRTUAL ASSISTANT TO SUPPORT INFORMATICS LESSON PLANNING

Tóm tắt

Preparing high school Informatics lesson plans is time-consuming, as teachers must meet the knowledge, competency, and quality requirements of the 2018 General Education Curriculum while organizing textbook content in accordance with the structure of Official Dispatch 5512. Large Language Models (LLMs) can assist teachers; however, hallucination – where models generate content that deviates from the curriculum – is the primary obstacle. This study develops a GraphRAG-based virtual assistant, combining Gemini with a Neo4j knowledge graph that encodes the high school Informatics curriculum and the Digital Competence Framework prescribed by Circular 02/2025/TT-BGDĐT and operationalized for school students in Official Dispatch 3456 (MOET, 2025b). The novelty of this research lies in a two-layer control mechanism: an immutable zone extracting data verbatim from the knowledge graph, and a creative zone for the model to flexibly design pedagogical activities. The pedagogical design of the system is justified through the Backward Design and constructive alignment frameworks, clearly defining the structural support role of the virtual assistant while preserving the teacher's pedagogical autonomy. A survey of 35 participants yielded an overall mean score of 4.52/5 (SD = 0.60; α = 0.898) across 22 criteria in four groups. A within-subject comparative experiment (n = 20) showed that the system was rated higher than a standalone LLM across all four core criteria (p < 0.05, Wilcoxon signed-rank test).