BackHome
Refresh: OffView Personal Schedule
Session Submission Summary
Share...

Direct link:

Tailored Instruction at Scale: Harnessing A.I. to Close Learning Gaps

Mon, March 24, 2:45 to 4:00pm, Palmer House, Floor: 3rd Floor, The Ashland Room

Group Submission Type: Formal Panel Session

Proposal

This panel will bring together experts from J-PAL, Pratham, Letrus, and Saga Education to explore the mechanisms behind successful tailored instruction interventions and the potential for AI to scale these approaches. The session will address how AI can support teachers more efficiently meet diverse student needs. Panelists will also discuss strategies for working with governments and school systems to scale tailored instruction effectively.

Despite significant increases in school enrollment globally, students are still not acquiring the skills they need to succeed in school and in their careers. Many education systems struggle to help students meet the curriculum standards they have set. For instance, over 75% of youth in Brazil fail to reach basic competence on Programme for International Student Assessment (PISA) tests, and 30% of U.S. 8th graders scored below basic proficiency in math on the 2017 National Assessment of Educational Progress (NAEP).

A major issue is that there are wide variations in students’ learning levels within the same grade. For example, a 2015 study in Delhi revealed that learning levels within grades 6 to 9 spanned five to six grade levels. Schools often incentivize teachers to cover the curriculum as prescribed, offering little flexibility to address gaps in students’ foundational knowledge. As a result, students struggle with advanced concepts without mastering the basics, further widening the learning gap.

Poverty exacerbates these challenges. Worldwide, a child’s family background – such as parental education, socioeconomic status, and home conditions – strongly influences learning outcomes. In the United States, for example, students from low-income families tend to start school behind their wealthier peers and struggle to catch up. However, this learning gap is not inevitable. PISA results highlight that high-performing school systems in countries like Canada, Finland, and Japan manage to provide quality education to all students, regardless of their income levels.

What interventions can better support teachers and students to close learning gaps? How can these interventions be scaled effectively?

Rigorous impact evaluations suggest that targeting instruction to students’ current learning levels, rather than the prescribed curriculum, can significantly boost student outcomes. Interventions like Teaching at the Right Level (TaRL) and tutoring programs focus on students with low performance, using tailored instruction to improve foundational skills. At their core, tailored instruction interventions engage students at their own learning pace, with more time on specific topics and faster teacher feedback.

For example, Pratham’s TaRL programs in India, tested across six randomized evaluations, improved test scores by 0.07 to 0.70 standard deviations, with low-performing students benefiting the most. The program assessed students’ basic literacy and numeracy skills and grouped them by learning level rather than age or grade, with instructors adjusting teaching strategies accordingly. Similar models have shown positive results in Chile and the United States.

Another tailored instruction model is individualized tutoring. Saga Education’s program for high school students features five main characteristics: regularly scheduled tutoring that takes place during the school day, personalized instruction in small groups of three students, supportive relationships with tutors, and a research-based curriculum. An evaluation in Chicago found that Saga’s model led to an extra one to two years of math learning, boosted GPAs by 0.58 points (out of 4.0), and reduced math failure rates by more than 50%. These outcomes suggest that even students facing significant barriers can improve their academic performance with targeted support.

However, scaling tailored instruction remains a challenge. Across different implementation models, teachers must accurately assess students’ learning levels and provide timely, customized support, which can be resource-intensive in already overburdened classrooms.

Advancements in technology are opening up new possibilities for scaling tailored instruction. AI-based tools, in particular, can assist teachers in effectively addressing diverse learning needs. One example comes from Brazil, where large class sizes have made it difficult for teachers to provide individualized writing instruction. In 2019, the state of Espírito Santo piloted an AI platform called Letrus to help students prepare for the national university entrance exam. Letrus provided students with instant feedback – offering suggestions, critiques, and statistics – that encouraged more frequent writing and deeper engagement. A randomized evaluation showed that the platform significantly improved students' writing skills and reduced the achievement gap between public and private school students. By automating parts of the feedback process, Letrus freed up teachers to focus on giving more personalized support, contributing to students’ writing improvements.

Panelists will engage audience members in discussions on leveraging AI to scale evidence-based tailored instruction interventions. They will explore key questions such as: How can tailored instruction models effectively close learning gaps across different contexts and education systems? What role can AI play in supporting teachers to tailor instruction efficiently, particularly in resource-constrained contexts? What are practical strategies for partnering with governments to implement and scale tailored instruction interventions?

Overall, the panel will convene innovative, evidence-driven organizations at different stages of AI adoption to think through the applications of AI in delivering tailored instruction.

Sub Unit

Chair

Individual Presentations