Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Computational Model of Inclusive Pedagogy: From Understanding to Application

Francesco Balzan, Pedro P. Santos, Maurizio Gabbrielli, Mahault Albarracin, Manuel Lopes

TL;DR

The paper tackles the gap between educational theory and AI by formalizing a co-adaptive teacher–student interaction (T-SI) framework that supports bidirectional learning in inclusive, one-to-many settings. It introduces a Bayesian, co-adaptive pedagogy model and tests five interaction modalities in a synthetic Guess Who game with three student groups of differing observability. Results show that combining adaptive teaching with active learning yields full inclusion and fastest mastery, surpassing unilateral strategies and highlighting the value of group-aware adaptation for equity. The framework provides a scalable sandbox for hypothesis generation, bridging ethnographic insights and scalable AI technologies, with implications for equitable AI in Education and beyond.

Abstract

Human education transcends mere knowledge transfer, it relies on co-adaptation dynamics -- the mutual adjustment of teaching and learning strategies between agents. Despite its centrality, computational models of co-adaptive teacher-student interactions (T-SI) remain underdeveloped. We argue that this gap impedes Educational Science in testing and scaling contextual insights across diverse settings, and limits the potential of Machine Learning systems, which struggle to emulate and adaptively support human learning processes. To address this, we present a computational T-SI model that integrates contextual insights on human education into a testable framework. We use the model to evaluate diverse T-SI strategies in a realistic synthetic classroom setting, simulating student groups with unequal access to sensory information. Results show that strategies incorporating co-adaptation principles (e.g., bidirectional agency) outperform unilateral approaches (i.e., where only the teacher or the student is active), improving the learning outcomes for all learning types. Beyond the testing and scaling of context-dependent educational insights, our model enables hypothesis generation in controlled yet adaptable environments. This work bridges non-computational theories of human education with scalable, inclusive AI in Education systems, providing a foundation for equitable technologies that dynamically adapt to learner needs.

A Computational Model of Inclusive Pedagogy: From Understanding to Application

TL;DR

The paper tackles the gap between educational theory and AI by formalizing a co-adaptive teacher–student interaction (T-SI) framework that supports bidirectional learning in inclusive, one-to-many settings. It introduces a Bayesian, co-adaptive pedagogy model and tests five interaction modalities in a synthetic Guess Who game with three student groups of differing observability. Results show that combining adaptive teaching with active learning yields full inclusion and fastest mastery, surpassing unilateral strategies and highlighting the value of group-aware adaptation for equity. The framework provides a scalable sandbox for hypothesis generation, bridging ethnographic insights and scalable AI technologies, with implications for equitable AI in Education and beyond.

Abstract

Human education transcends mere knowledge transfer, it relies on co-adaptation dynamics -- the mutual adjustment of teaching and learning strategies between agents. Despite its centrality, computational models of co-adaptive teacher-student interactions (T-SI) remain underdeveloped. We argue that this gap impedes Educational Science in testing and scaling contextual insights across diverse settings, and limits the potential of Machine Learning systems, which struggle to emulate and adaptively support human learning processes. To address this, we present a computational T-SI model that integrates contextual insights on human education into a testable framework. We use the model to evaluate diverse T-SI strategies in a realistic synthetic classroom setting, simulating student groups with unequal access to sensory information. Results show that strategies incorporating co-adaptation principles (e.g., bidirectional agency) outperform unilateral approaches (i.e., where only the teacher or the student is active), improving the learning outcomes for all learning types. Beyond the testing and scaling of context-dependent educational insights, our model enables hypothesis generation in controlled yet adaptable environments. This work bridges non-computational theories of human education with scalable, inclusive AI in Education systems, providing a foundation for equitable technologies that dynamically adapt to learner needs.
Paper Structure (41 sections, 3 equations, 3 figures)

This paper contains 41 sections, 3 equations, 3 figures.

Table of Contents

  1. Introduction
  2. Computational approaches to T-SI
  3. Non-Computational insights on T-SI
  4. The computational approach
  5. Computational models of Active Teaching
  6. Computational models of Active Learning
  7. Computational models of Active Teaching and Learning
  8. Computational Model of Co-adaptive Pedagogy
  9. Experimental Scenario: The Collaborative "Guess Who" Game
  10. Students' observability
  11. Groups of students
  12. Objective and research question
  13. Interaction Modalities
  14. Active Learning: Student-Driven Questions
  15. Active teaching: Teacher-Guided Clues
  16. ...and 26 more sections

Figures (3)

  • Figure 1: Illustration of the teacher-student interaction setting considered. In the illustration, we consider a set of 15 students divided among three groups $\{g_1,g_2,g_3\}$, each composed of 5 students. We consider three characters, $\{$Alex, Mary, and John$\}$, and the set of traits used to identify each of the characters is $\{\text{Glasses}, \text{Brown hair}, \text{Hats}\}$. The left top table displays the value that each trait takes for each of the characters: a value of one indicates that such a trait is present/active for the respective character; on the other hand, a value of zero means that the trait is not present/active. The set of types of the students is $\{$, $\}$. The left bottom tables display the description of the characters from the perspective of students of types and . As can be seen, students of type are not able to identify whether the characters have brown hair, and students of type are not able to identify whether the characters are wearing a hat or not.
  • Figure 2: Percentage of students correctly identifying the target character across interaction steps. Shaded regions show the 95% confidence intervals over 100 simulations. Left: Initial interactions (teacher has no prior group experience). Middle: After 20 group interactions (moderate adaptation). Right: After 40 interactions (experienced teacher). Adaptive teaching + active learning (orange) achieves full inclusion.
  • Figure 3: Illustration of the teacher-student interaction setting considered. In the illustration, we consider a set of 15 students, $\mathcal{S}$, divided among three groups $\mathcal{G} = \{g_1,g_2,g_3\}$, each composed of 5 students. The set of features is $\Phi = \{\phi_1, \phi_2, \phi_3\}$ and the set of concepts is $\mathcal{Y}=\{y_1,y_2,y_3\}$. At the current iteration, the teacher aims to teach the concept $y^* = y_3$ to all students. The left top table displays the value that each feature in $\Phi$ assigns to each of the concepts in $\mathcal{Y}$. The set of types of the students is $\mathcal{T} = \{$, $\}$. The left bottom tables display the description of the concepts from the perspective of students of types and . As can be seen, students of type are not able to identify the values of feature $\phi_2$, and students of type are not able to identify the values of feature $\phi_3$.