Examining the Use of an AI-Powered Teacher Orchestration Tool at Scale

There is an increasing opportunity for AI-supported teacher-student orchestration. Preliminary evidence in small studies suggests that AI that better supports such coordination could lead to substantial student learning gains but much is unknown about how such orchestration might work at scale. In this work we focus on an existing tool, widely available to teachers as part of a commonly used math platform, to gain insights into how teachers perceive this tool and how they use a feature which allows them to mark when and why they help particular students as those students work on the software. Our teacher survey reveals that many teachers do use the tool's suggestion to inform which students to support, and our quantitative analysis of log files shows that when marking a student as helped, teachers most often report providing encouragement. Providing encouragement is far more frequent than providing direct math instruction, though when students are highlighted as being likely to fail the current section, teachers more frequently mark that they provided direct math help. Our work helps showcase the existing use and potential new directions for AI-supported teacher-student orchestration at scale.