Modeling And Optimizing Learning Trajectories in Remedial Education: Evidence from Teaching at The Right Level in Madagascar

Abstract Book of the 10th World Conference on Research in Education

Year: 2026

[PDF]

Modeling And Optimizing Learning Trajectories in Remedial Education: Evidence from Teaching at The Right Level in Madagascar

Razafindrafara Elysa Marie Alfredine, Razafinirina Mahefa, Nguala Jean Berky, Raherinirina Angelo

ABSTRACT:

Teaching at the Right Level (TaRL) is a widely used remedial education approach in low- and middle-income countries, designed to address learning heterogeneity. While its overall effectiveness is well documented, less is known about the internal learning dynamics and the sequencing of pedagogical activities that shape student progression. Learning in TaRL can be understood as a complex system, where students, instructional practices, and time interact in non-linear and uncertain ways. Using data from the implementation of TaRL in Madagascar, this study proposes an integrative modeling framework to analyze and optimize learning trajectories within the program. Student progression across discrete competency levels is modeled using absorbing Markov chains to estimate learning time and identify structural bottlenecks. Pedagogical sequencing is then formalized as a Markov Decision Process (MDP) to derive optimal instructional policies minimizing expected time to mastery. To account for uncertainty in observed performance, latent competency states are inferred using Hidden Markov Models (HMM), enabling a belief-based decision perspective. Empirical instructional strategies are additionally extracted from observed classroom sequences using supervised machine learning. Findings highlight contrasting learning patterns across domains: literacy trajectories are largely progressive, whereas numeracy learning exhibits significant frictions, including stagnation and regression, particularly at intermediate levels. Comparisons between optimal model-based policies and observed practices reveal systematic gaps, suggesting opportunities for improving instructional sequencing. By combining learning time analysis, stochastic modeling, and decision theory, this research offers practical insights for strengthening remedial education programs. The framework supports evidence-informed pedagogical decisions and contributes to understanding how durable and equitable learning pathways can be designed in complex educational contexts, with particular relevance for education systems in Madagascar and similar contexts.

Keywords: Teaching At The Right Level; Learning Time; Learning Trajectories; Educational Modeling; Instructional Decision-Making