Why AI Struggles to Think Like a Student

Drawing on the work of Lee Shulman and a recent interview with researcher Andrew Lan, we look at why AI can generate content and imitate pedagogy, yet still struggle to understand how real students misunderstand, make mistakes, and learn. We also consider what this means for evaluating EdTech tools, teaching AI literacy, and the future potential of simulated student agents in instructional planning.Topics covered: • What pedagogical content knowledge (PCK) is, and why it matters so much in teaching • Why AI still struggles to simulate authentic student thinking and misconceptions • How this helps explain the “jagged frontier” of AI in education • Why many AI-powered education tools may underdeliver in real classroom settings • The difference between knowing content, knowing pedagogy, and knowing how students learn specific content • Why current approaches to AI literacy may focus too much on tools and not enough on underlying understanding • How simulated student agents could eventually help teachers test lessons and prompts before instruction • Why teacher expertise remains difficult to replicate and especially valuable in the age of AISource:https://the-learning-agency.com/the-cutting-ed/article/5-questions-with-andrew-lan/