Tutoring is a hot topic this summer as schools and parents search for ways to help children rebound from COVID learning loss. From the call for a national tutoring corp last January to the Biden Administration’s educational investments in the American Families Plan in April, the nation may be at a turning point on how it might implement and scale up tutoring for kids who need it most.
But one thing people are not talking much about is the opportunity — right now — to learn something really important: How to improve and share the components of so-called intelligent tutoring systems. Learning how to effectively design and operate these systems is a key step to bringing efficient, effective tutoring to every student who needs it.
Renowned education researcher Robert Slavin’s blueprint for a tutoring corp, combined with some solid meta-analyses, has suggested that tutoring is effective, but expensive. High-dosage tutoring costs so much because it requires three in-person tutoring sessions per week (limited to two or three students per tutor). That translates into at least $3,500 per student, per subject per year. Providing tutoring only to those students in Title 1 schools, the endeavor would cost roughly $130 billion annually.
As part of the American Families Plan, the Biden Administration is directing $40 billion to supplemental educational services, funds which have been traditionally used for schools to pay for services like tutor.com. This infusion of federal funding is a tremendous opportunity to learn how the ed-tech sector can improve intelligent tutoring systems, expand their reach and thereby bring down the costs of high-quality computer tutoring.
Today, millions of students are doing their independent math practice on educational platforms — platforms that are key to lowering costs. When funding tutoring studies, philanthropic foundations usually choose to work with platforms that students are already using. The next logical step is to design a less expensive online variation of high-dosage tutoring, where students receive personalized help via a host platform such as ASSISTments, McGraw-Hill’s ALEKS, Carnegie Learning’s MATHia, etc.
If a platform is going to launch and host tutoring sessions, it must ensure, through measurement, that federal investments yield a lasting return — knowledge that informs how such tutoring can be made better and more efficient. If the United States plans to spend $40 billion on tutoring, it should also learn more deeply about what works in tutoring so that when this rush of COVID funding is spent, the host platform that students interact with will have been made better. And can it be determined which tutors were more effective by using students’ post-session performance on other assessments?
Platforms should do more than just serve as the connection between tutor and students. Why not plan ahead and set up an infrastructure that enables us to study what works and what doesn’t? How might platforms accomplish all this? If tutors are being randomly assigned to many students, it is fairly easy to measure their effectiveness by seeing how students perform on similar problems, post-session, while ensuring student privacy.
For example, if a tutoring provider recognizes a few tutors are doing a great job providing motivational messages to the students doing computer-supported math homework, these platforms could study what their most successful human tutors are saying to motivate their students. They could then organize randomized controlled experiments with larger groups of students to test whether or not their messages cause better outcomes. This second step means that for every student a tutor effectively helps, thousands more could receive the same support as platforms scale up these ideas. This one example is meant to show how funding for such a project can help many more students, but there are so many possibilities: like conducting similar research on what explanations cause better learning, for example.
It’s an exciting time to be working in the tutoring space of learning technology. If the sector can expand access to personalized tutoring to diverse and varied socioeconomic populations of students, while also allowing learning technology researchers to study the sessions, our knowledge of tutoring could grow deeply and quickly. And if used carefully, the current wealth of funding opportunities could lead to future proposals for computer-based versions of tutoring that can be tested at scale in a manner consistent with IES’s Standards for Excellence in Education Research (SEER).
The type of ideas and research I’m calling for are not new. In fact, IES has a dedicated grant line that supports platforms to engage in just this kind of research. Let’s use this post-COVID funding opportunity to advance our knowledge forward and expand personalized tutoring to even more students across the country.