To improve ed tech, focus on feedback

Automated feedback can supplement in-person instruction by making teachers' jobs easier while personalizing student experience.
Neil Heffernan is professor of computer science at Worcester Polytechnic Institute.
Neil Heffernan is professor of computer science at Worcester Polytechnic Institute.

The pandemic seriously tested the state of educational technology. In a lot of respects, it was found wanting. Teachers complained that, even before the pandemic, ed tech was often rolled out without their input and without sufficient training. Students complained that ed tech was impersonal and confusing. And parents grew frustrated with misfiring apps, dropped connections, and lost learning.

As students — hopefully — finally return to their classrooms, some are calling for a serious reappraisal of the role of ed tech in students’ education

There is certainly a great deal of room for improvement, but as we rethink the role of ed tech in a post-pandemic world, it’s worth focusing on what works well in the field and building on it. In particular, I want to discuss the promise of automated feedback, which can supplement in-person instruction by making teachers’ jobs easier while personalizing student experience.

One of the most common complaints about ed tech – especially in subjects like math, where pencil and paper still dominate – is that it lacks the rich, in-depth interactive quality of one-on-one teaching. Without the constant feedback provided by the best teachers and tutors, students get lost, lose interest, or spin their wheels in frustration.

It doesn’t have to be this way. Lots of ed tech products are integrating automated tools to improve student engagement and success while making grading and feedback easier for teachers.

Edtech platforms and tools, like Carnegie Learning and Zearn, have shown success with formative feedback, giving students immediate, scaffolded feedback to support improved student performance. Although the earliest successes with integrated feedback have come with math platforms, automated feedback on tasks like writing is rapidly evolving as well, leveraging advances in natural language processing technology.

In order to be successful, feedback must do more than simply identify an error or give the student an answer they’re missing. The best feedback tools recognize there is more than one way to come to a solution; it can offer students and teachers multiple feedback types; and it can adjust depending on the individual student’s particular struggle and context. In other words, it must respond to students’ inputs more like a human tutor might.

The next steps for automated feedback, therefore, involve testing feedback to determine what works best for students and teachers. ASSISTments, a math homework platform I developed at the Worcester Polytechnic Institute, is piloting a feedback tool, called QUICK-Comments that allows teachers to quickly select from a number of different recommended feedback messages selected by AI. The AI improves its recommendations over time, learning through iteration what works best for students and teachers.

Moreover, feedback can become more responsive if we leverage the best human tutors and integrate them into automated feedback tools. One thing good human tutors consistently do well is to detect frustration or disengagement in their pupils and adjust instruction accordingly. High-cost physical sensors can detect student affect, but sophisticated lower-cost detectors that respond to students’ actions within the software are being developed as well. As we continue to improve our understanding of the interaction between students and tutoring programs, so will the feedback these systems can offer students to keep them engaged, challenged, and learning.

Finally, we should also begin to work on supporting tutors with AI the ways we have been supporting teachers. Through programs like Education, Innovation and Research grant, the Department of Education is helping researchers and technologists like my team explore enhancing in-person tutoring to help remediate COVID-related learning loss.

After the pandemic, ed tech can resume doing what it does best, supporting human teachers and tutors rather than trying to stand in for them. If we learn the right lessons from the pandemic, the improvements to instruction and student learning can be great.

Neil Heffernan is the William Smith Dean’s Professor of Computer Science at Worcester Polytechnic Institute. He developed ASSISTments, a web-based learning platform, with his wife Cristina Heffernan.

More from DA

Most Popular