An Agent-based Classroom Lessons Model and Simulation

Ingram, Fred and Brooks, Roger and Williams, Richard (2020) An Agent-based Classroom Lessons Model and Simulation. PhD thesis, Lancaster University.

Text (2020ingramphd)
2020ingramphd.pdf - Published Version
Download (9MB)

Abstract

In education contexts, agent-based simulations have been used at all levels, from the individual to the classroom to the state, to explore individual-level interactions and often to provide conclusions and explanations that could support decision-making. The aim of this research was to develop an empirically-grounded classroom lessons behaviour model and explore the theoretical consequences of the model design and the modelling assumptions, to understand the mechanisms of lesson interactions and to assess whether the results and explanations were plausible and realistic. The model was to be as realistic (not simplistic) as possible, addressing the limitations found in current lesson-related simulations and extending the use of agent-based simulations in classroom research to full-lessons with full-classes. A typical empirically-driven agent-based modelling and simulation methodology was followed, incorporating an investigative case study at a UK secondary school. A comprehensive agent-based model of classroom lessons was formulated and a lesson event recording tool was developed and used to record a wide range of student and teacher activities. These data were used to calibrate and validate the simulation model. Three agent types were incorporated: students, teacher and teaching assistant. Agent decision-making was modelled using conventional production rules (one set for each agent type) that integrated the influences from, in order of significance, the lesson plan (which specifies the desired behaviour and is enacted by the teacher), the current circumstances (who is doing what and where) and, for students only, their historical empirical activity state frequencies. The simulation model was validated with the help of experienced teachers, who considered that it embodied plausible theories of classroom behaviours. It was seen to generate plausible causal agent-level explanations for some important lesson dynamics and therefore judged to be a useful lesson analytics and decision-support tool, enabling educators to explore the consequences of a range of lesson interventions.