Introducing generative AI with Markov Chains: Gendered patterns of competence in English Language Arts classrooms
Daria Smyslova, Shiyan Jiang, Carolyn Penstein Rosé, Rebecca Ellis, Jie Chao, Qiankun Li · The Journal of Educational Research · 2025
This study examined an intervention designed to foster high school students’ AI literacy through foundational text generation models. Using Markov Chains as an entry point, the study supported students’ understanding of predictive modeling and the probabilistic nature of AI-generated text. Using a mixed-methods approach, pre- and post-assessment showed significant gains in students’ self-reported competence and understanding of AI text generation, while qualitative analysis highlighted improvements in recognizing how predictive models generate text sequences. However, findings suggested that while students developed a foundational understanding, they faced challenges in extending this knowledge to more advanced AI systems. Some misconceptions also persisted, including the belief that AI-generated text is random rather than probabilistic. Also, female students tended to underestimate their competence despite slightly higher learning gains. These findings underscore the need for structured scaffolding to bridge foundational and advanced AI concepts, ensuring students develop both technical understanding and critical evaluation skills.