A minimax optimization framework for adaptive teaching strategies in multilingual classrooms

Authors

  • Sherzod Yuvashov Author
  • Ravqat Masharipov Author
  • Sodiq Majidov Author
  • Bexzod Isxakov Author
  • Sanobarkhon Inomkhujaeva Author
  • Yanglish Kosimova Author
  • Xurshid Tursunov Author

Keywords:

Minimax optimization, multilingual education, adaptive teaching, robust decision making, translanguaging, linear programming, multiplicative weights.

Abstract

Multilingual classes have very diverse levels of language abilities of students in the language of instruction. Teacher is required to make his choice between different types of teaching strategies without knowledge of exact numbers of the students of each linguistic profile in a particular unit, and these numbers vary in time during a semester. Almost all researches in this field focus on analyzing particular strategies or optimizing for the average class profile, and it results in suboptimal learning outcomes for the most poorly-performing students when the class is different from the assumed one. In this paper, we propose minimax framework for selection of teaching strategies. Namely, we treat this problem as a zero-sum game where the teacher chooses her mixed strategy, i.e. a proportion of time allocated to each strategy. It is shown that the problem can be reduced to one linear program via duality, and we incorporate an easy multiplicative-weights scheme that enables the teacher to adjust the strategy week to week based on formative feedback. The proposed framework is evaluated in a simulation setting using five different language student profiles and six teaching approaches. For an uncertainty radius of 0.10, the minimax mix (around 48% of scaffolding approach and 52% of translanguaging approach) increased the minimum gain from 0.516 to 0.536 compared with the average optimal approach, and increased the gain of the weaker learner from 0.22 to 0.36 with a cost of 0.024 in the nominal case. For an unanticipated shift in class composition outside the assumed range, the benefit increased further. The minimax initialised adaptive variant performed the best on average weekly gain for a mid-term shift in class composition. The payoffs are modelling assumptions, so the findings only describe the behaviour of the framework.

Downloads

Download data is not yet available.

Downloads

Published

2026-09-22

How to Cite

A minimax optimization framework for adaptive teaching strategies in multilingual classrooms. (2026). Minimax Theory and Its Applications, 11(2s), 1-11. https://journalmta.com/index.php/jmta/article/view/392