When AI Took the Parsing, the Teacher Took the Argument

In this episode, we explore a fascinating (and counterintuitive) finding from a large-scale study of AI writing tools in Brazilian high schools: adding human graders to AI feedback didn’t improve student outcomes—but that’s not the real story.We challenge the common narrative about AI “doing the teaching” and instead ask a more important question: what happens when AI handles the routine work so teachers can focus on what matters most?In this episode, we explore:How an AI writing tool improved student outcomes—even without human gradersWhy the biggest gains showed up in complex thinking, not surface-level writingThe concept of “task reallocation” and why it matters more than automationHow teachers used AI to shift toward more meaningful one-on-one conversationsWhy more writing + more feedback didn’t require more teacher timeThe critical role of sequencingWhy the lowest-performing students saw no benefitThe limits of AI as a “last mile” solution for feedbackWhat educators should really ask before adopting any AI toolSource:https://www.povertyactionlab.org/sites/default/files/research-paper/working-paper_5806_Artificial-Intelligence-Teacher-Tasks-Pedagogy_Brazil_Feb2021_0.pdf