ע 123 עידית גת , מאיה אושר , מירי ברק ספר הכנס העשרים ואחד לחקר חדשנות וטכנולוגיות למידה ע"ש צ'ייס: האדם הלומד בעידן הדיגיטלי א' בלאו, ד' אולניק - שמש, נ' גרי, א' כספי, י' סידי, י' עשת - אלקלעי, י' קלמן ו נ' ברנדל )עורכים(, רעננה: האוניברסיטה הפתוחה בין דיוק מדעי לחדשנות טכנולוגית: הערכת תוכן שנוצר באמצעות בינה מלאכותית יוצרת )מאמר קצר( Between scientific accuracy and technological innovation: Evaluation of AI-generated content (Short paper) Abstract The growing integration of generative artificial intelligence (GenAI) technologies into science education underscores the need for systematic, critical evaluation of AIgenerated content, which may contain scientific inaccuracies and reinforce misconceptions. This mixed-methods study examined how science teachers evaluate content generated by ChatGPT and how they perceive the integration of such evaluative practices within their teaching. Sixty middle-school science teachers participated in a professional development (PD) workshop at the Technion. During the workshop, teachers used ChatGPT to generate questions at different cognitive levels and corresponding answers, then evaluated them drawing on their pedagogical content knowledge for scientific accuracy, linguistic clarity, and curriculum alignment. A dualanalytic approach combined quantitative analyses of numerical ratings with qualitative analyses of written explanations and reflections. The findings revealed that lower-order thinking questions received consistently high evaluations, whereas higher-order thinking questions were perceived as containing conceptual gaps and vague formulations. Nonetheless, teachers recognized the pedagogical value of such questions for fostering reflective thinking and connecting the content to everyday contexts. Participants' reflections emphasized that structured evaluation of GenAI outputs can enhance instruction by promoting disciplinary discourse, broadening assessment practices, bridging differences in teaching experience, and redefining the teacher's role as an "evaluator." The findings highlight the importance of incorporating critical evaluation practices of AI-generated content into PD programs for science מירי ברק הטכניון – מכון טכנולוגי לישראל bmiriam@ed.technion.ac.il מאיה אושר HIT מכון טכנולוגי חולון הטכניון – מכון טכנולוגי לישראל mayau@ed.technion.ac.il עידית גת הטכניון – מכון טכנולוגי לישראל Idit.gat@campus.technion.ac.il Idit Gat Technion – Israel Institute of Technology Idit.gat@campus.technion.ac.il Maya Usher HIT Holon Institute of Technology Technion – Israel Institute of Technology mayau@technion.ac.il Miri Barak Technion – Israel Institute of Technology bmiriam@ed.technion.ac.il
RkJQdWJsaXNoZXIy Mjk0MjAwOQ==