Maayan Shay Sayag, Ina Blau, Orit Avidov-Ungar E53 associated with different levels of GenAI integration (RQ2), analyzing both temporal progression and implementation approaches. The analysis reveals that while teachers engaged with GenAI across multiple pedagogical domains, the critical differentiating factor for achieving transformative integration was not time elapsed since training, but rather the level of implementation—specifically, whether activities involved direct student engagement. The Pedagogical Landscape: Mapping GenAI Integration To understand how teachers integrated GenAI tools into their professional practice, we mapped their pedagogical activities across three core domains: Teaching, Learning, and Assessment. This classification provides a comprehensive view of how GenAI integration manifests across the full spectrum of pedagogical practice. Table 1 presents the distribution of all coded activities across these pedagogical domains and SAMR levels. Table 1. Distribution of Activities by Pedagogical Domain and SAMR Level Pedagogical Domain Substitution Augmentation Modification Redefinition Total Teaching 10 48 8 8 74 Learning 2 14 43 9 68 Assessment 5 21 11 2 39 The distribution reveals distinct integration patterns across pedagogical domains and SAMR levels. Substitution activities (9.4% of total activities) represented direct replacement of existing practices without functional improvements. In Teaching contexts, teachers used AI to generate lesson plans that replicated traditional planning processes, though one biology teacher noted the AI "gave me a truly nice lesson plan with few ideas I hadn't thought of" (T9B-R1). In Assessment contexts, AIgenerated test questions often required substantial human revision to achieve appropriate cognitive complexity. When students used AI as a direct search engine replacement in Learning contexts, outcomes were frequently problematic, with one teacher reporting "shocking" results (T5B-R1). These limitations align with concerns about difficulties in translating AI capabilities into pedagogically sound implementations (Ding et al., 2024; Kong & Yang, 2024). Augmentation activities constituted the largest category (45.9% of total activities), representing functional improvements over existing practices. Teaching activities concentrated primarily at this level (64.9% of Teaching activities), suggesting teachers used GenAI predominantly to enhance existing instructional practices. Teachers leveraged AI for enhanced brainstorming, with one civics teacher describing: "I asked it to give me ideas for a civics text influenced by other fields... I simply got a whole world of options" (T11C-R1). In Learning contexts, students used AI to overcome language barriers when reading academic articles. As one teacher noted, students would "find the article, put it in ChatPDF and ask the tool to explain methodologies and findings," adding that this approach "really saved time" (T5B-R1). In Assessment contexts, AI enabled efficient creation of practice materials, such as converting informational text into fill-in-the-blank exercises (T5B-R1). These applications enhanced efficiency but did not fundamentally redesign learning tasks, consistent with findings that most GenAI activities concentrate at enhancement levels (Shamir Inbal et al., 2024; Jiménez-García et al., 2024). Modification activities (34.3% of total activities) involved significant task redesign enabled by AI capabilities. A striking pattern emerged: Learning activities, where students directly engaged with AI tools, showed marked concentration at this level (63.2% of Learning activities), indicating more transformative integration when AI reached students directly. Students shifted from content creation to critical evaluation of AI outputs. One teacher designed an activity where students would
RkJQdWJsaXNoZXIy Mjk0MjAwOQ==