When the Rubber Meets the Road: Lessons from the In-School Adventures of an Automated Reading Tutor That Listens1

tions by automating them. Automated intervention ensures treatment fidelity over time and across multiple schools and settings in a way that human instruction cannot. Automated tutoring is consistent whether studied in smallor large-scale implementations—a key property for scalability. That is, contextual factors may affect the amount of tutoring (implementation intensity), but not the intervention itself. This point is a bit more subtle than it may seem, because automated interventions can and do respond differentially to student behavior and teacher input. However, the intervention design remains intact. In contrast, human interventions can be transformed when implemented (Coburn 2001; Elmore 1996; Hoffman, McCarthey, and Elliot et al. 1998; Spillane and Jennings 1997; Spillane and Zeuli 1999; Stein, Grover, and Henningsen 1996), with teachers altering interventions in ways that can negate the very components responsible for strong effects in experimental trials. Lasting educational change is difficult to achieve because it must change not only educational materials but teacher practice and beliefs (Fullan 2001). Other projects funded by IERI (IERI 2002) and the U.S. Department of Education focus on trying to fundamentally change the ways teachers teach reading. However, the more a new practice differs from an existing practice, the harder it is to implement (Cohen and Ball 2001). The main changes required to implement an automated intervention are to schedule and value its usage. These requirements are challenging, but straightforward compared to deeper transformations in how teachers teach. Ideally, the technology is installed, students use it, and they learn.