Since ChatGPT launched in 2022, teachers and administrators have had the challenge of using generative Artificial Intelligence (AI) effectively while minimizing the negative consequences of its presence in schools. Today, AI-enabled lesson plan generators such as Diffit and MagicSchool AI are widely available to teachers, but no research has explored the quality of the lesson plans these tools produce. This leaves administrators with few guidelines to support teachers’ effective use of these products. This study addresses this need by using qualitative content analysis methods to compare the outputs from these AI lesson plan generators. We analyzed the extent to which these outputs matched an evidence-based lesson planning framework: Universal Design for Learning/Universal Design for Transition (UDL/UDT). Results show that AI-generated lesson plans have minimal alignment with UDL/UDT and require teacher-initiated modifications to successfully support learners with diverse needs. These findings suggest that the products of AI lesson plan generators may lack the specificity needed to provide novel ideas for experienced teachers, although they may be useful for teachers in their inductive years, or occasionally for experienced teachers. We conclude with suggestions to align AI-produced lesson plans with the UDL/UDT tenets of engagement, representation, and expression.
Read full abstract