Yogev Shani, Idit Adler E87 approach designed to support cognitive and metacognitive learning through guided experiences and externalization of thinking. CAM involves several stages: modeling, where an expert models a task; coaching, which serves to direct the trainee’s attention to a previously unnoticed aspect of the task; scaffolding, where the expert provides help that enables the trainee to carry out the task; articulation, which involves prompting trainees to articulate their knowledge, reasoning, or problemsolving processes; reflection, in which trainees compare their problem-solving approaches with those of an expert; and exploration, which encourages trainees to engage in independent problemsolving (Collins et al., 1989, 1991; Dennen & Burner, 2008). However, the small facilitator-to-learner ratio required is a barrier to its large-scale implementation (Collins, 2006). Generative Artificial Intelligence (GenAI) may address this challenge by engaging PSTs in chatbot-led reflective discourse like that of a human mentor. The detect-diagnose-act framework, where AI receives data, diagnoses a learner’s current state, and translates the diagnosis into meaningful pedagogical actions, captures the basic functioning of AI in education (Molenaar, 2022). This study examined the potential of TeachPal, a GenAI-based chatbot designed to trigger teachers’ reflection on SRL instructional practices through personalized CAM-based feedback. TeachPal was implemented in a simulationbased course which alternated between human- and chatbot-led reflective discussions. We examined chatbot-led interactions between TeachPal and PSTs, and asked: 1. What characterizes PSTs’ discourse with TeachPal? 2. What characterizes TeachPal’s discourse with the PSTs? Methodology The study took place in a course at a major university. The course introduced PSTs to SRL through simulation-based learning. PSTs completed collaborative simulated scenario tasks (CSTs) and engaged in CAM-based discussions in class, and then used TeachPal, a GenAI-based chatbot, as an at-home reflective coach. TeachPal provided scenario-specific feedback aligned with CAM stages, supporting PSTs’ analysis and improvement of SRL-supportive instructional practices. Fifteen PSTs, all prepared to teach middle- or high-school science and who held at least a bachelor’s degree in this field, participated in this study. The PSTs’ discussions with TeachPal (n=60) were analyzed based on the methodology described by Robertson et al. (2020). Each discussion was first divided into episodes according to CAM stages (Table 1). Table 1. Discussion Episodes CAM Episode / Teaching model Definition Phase 1: Coaching TeachPal evaluates the PST’s instructional practices based on Opp4SRL, rates the levels of the different practices PSTs used, and draws their attention to how they can better leverage opportunities for SRL. Phase 2: Scaffolding TeachPal suggests that PSTs use higher-level instructional practices that promote SRL, tailored to the specific scenario. Phase 3: Modeling TeachPal conducts a renewed simulation of the situation in which the PST participated, using SRL-supporting instructional practices at a higher-level than those used by the PST. TeachPal conducts a reflection on the simulated scenario.
RkJQdWJsaXNoZXIy Mjk0MjAwOQ==