Aligning instruction with student needs: Evaluating instructional levelling strategies in LMICs with significant heterogeneity

• In Event: The use of assessments for student learning and educational improvement

Proposal

In many lower and middle-income countries (LMICs), there is a persistent misalignment between students' actual learning levels and the demands placed upon them by national curricula and the corresponding learning materials (Atuhurra & Kaffenberger, 2022; Glewwe, Kremer, & Moulin, 2009). This system-level misalignment leads to ineffective learning for many students, exacerbating educational inequities and perpetuating learning poverty (Pritchett & Beatty, 2015). Lev Vygotsky’s Zone of Proximal Development (ZPD) emphasises the importance of aligning instruction just beyond a student's current abilities (Vygotsky, 1978), yet many educational systems fail to achieve this balance. The gap between the curriculum and students’ abilities reveals the need for realigning educational content to match learning levels, which could significantly improve learning outcomes, especially for students most at risk of falling behind.
Empirical evidence supports the promise of targeted instruction, as demonstrated by computer-adaptive learning models and differentiated classroom approaches. However, these technology-driven solutions often remain inaccessible at scale in LMICs. Non-technology-based interventions, such as tutoring and tailored instruction, have shown similar potential in addressing diverse learning needs. Many initiatives in LMICs demonstrate the feasibility and impact of aligning instruction at scale in improving foundational learning. Despite these successes, educational systems often set curriculum expectations far above the learning levels of most students, sometimes targeting the median student in only the best-case scenarios, thereby overlooking the significant heterogeneity both within and across classrooms and schools. As research increasingly recognises these various forms of heterogeneity, there is growing support for more differentiated approaches that consider differences across school-grade combinations.

This work contributes to the ongoing discussion by examining various approaches to aligning instruction, identifying the most appropriate approaches to serve the needs of most students, and exploring the trade-offs associated with these strategies. We propose data-driven methods to evaluate the effectiveness of these different approaches, while accounting for heterogeneity at multiple levels and offer a framework for education stakeholders—particularly in contexts where technology-enabled solutions can help address the complexity of learning disparities.


Research Questions
How can instructional alignment be effectively tailored to address the diverse learning needs of students?
How do these instruction alignment approaches vary depending on the heterogeneity within classrooms and across schools for a given grade?
Is there a broader, more comprehensive framework for thinking about instructional alignment or levelling i.e., are there any methods that can be used to evaluate the effectiveness of such levelling approaches? How well do these approaches address the gap between students' abilities and instructional content?

Methods and Data
We draw on pupil-level data from 1,443 schools in Kwara, Nigeria, and 760 schools in Rwanda, covering more than 600,000 children. The data includes assessments of foundational literacy, specifically oral reading fluency, and numeracy across a broad range of skills.

To evaluate the different approaches to aligning instruction, we focus on two key goals. The first goals is to assess whether the instructional approach reaches the most number of students with the right instructional level, measured by the share of students that are levelled appropriately i.e., the hypothetical classroom instructional level exactly matches with the instructional level of the pupil or in other words, the instructional content meets them where they are. The second goal is to minimize the distance or the gap between each student’s learning level, and the hypothetical classroom instructional level. This is operationalized by assigning a number to each learning level (A=1, B=2, C=3 etc.), and then finding the difference in absolute terms between a student’s llevel and the hypothetical classroom instruction level, and then taking the average. The higher the number is, the worse it is from an instruction point of view.

We test various levelling approaches, including targeting the median learning level, the level with the most students, and a combination of both. These approaches can be applied at the system level or more finely differentiated by school and grade, with their effectiveness determined by the evaluation methods described above.

Preliminary Findings
Our work suggests the following -
The approach to differentiate instruction based on the median student within a school and grade tends to be more effective than applying a system-wide median for the same grade. This approach not only aligns instruction more closely with a larger share of students but also minimises the overall distance between individual students' learning levels and the hypothesised classroom instructional level.


As expected, the approach that aligns instruction to the level with the most students in a school and grade, matches the learning level of the largest number of students. However, in classrooms with significant learning heterogeneity, this method may not necessarily minimise the average distance between the instructional level and the varying learning levels of individual students.


Hence, the choice of levelling approach should depend on the instructional goal—whether to match the learning level of the greatest number of students or to reduce the relative gap between students' learning levels and the instructional content.

Contribution
The contribution of our work lies in offering new ways for evaluating instructional levelling approaches that account for the complex heterogeneity within education systems, particularly in lower and middle-income countries. By examining both within-classroom and between-school disparities, we challenge the traditional focus on system-wide median levels, proposing differentiated strategies that cater to school and grade-specific learning needs. Our work introduces two measures for assessing levelling effectiveness, which can guide more tailored and data-driven instructional interventions. Additionally, we highlight how technology-enabled solutions can be leveraged to better address this heterogeneity, particularly in contexts where non-technology methods fall short. This work extends the literature on levelling by providing actionable insights for education programme designers to improve instructional alignment and equity.

Authors

• Pranav Bhargava, NewGlobe Education

• Sylvester Mchihi, NewGlobe Education

• Keuna Cho, NewGlobe Education

• Priscilla Lu