FROM NEURONS TO CLASSROOMS: MULTI-SCALE MODELS OF LEARNING
This dissertation examines learning as an adaptive process across three distinct scales: educational frameworks, social-cognitive development, and neural computation. Through three complementary studies, we investigate how learning emerges from structured information environments at different levels of organization. First, we develop and validate a semi-automated Response to Intervention framework for early mathematics education, demonstrating significant improvements in arithmetic fluency and number concept understanding in first-grade students, even in resource-constrained educational environments. Second, we analyze the relationship between cognitive profiles and multilayered social networks among 5,804 students aged 6-15, revealing that while cognitive abilities organize along a single achievement dimension, social relationships show increasing differentiation throughout development, with children as young as 8 distinguishing between academic and recreational relationships. Third, we create a fully connectome-constrained model of the Drosophila melanogaster visual system, demonstrating that approximate numerical discrimination emerges naturally from verified neural architecture, with performance showing Weber ratio dependence characteristic of the Approximate Number System across species. Collectively, these findings illuminate how adaptive systems—from neurons to children in classrooms—extract and utilize structured information from their environments. This multiscale perspective enhances our understanding of fundamental learning mechanisms, with implications for educational practice, developmental psychology, and computational neuroscience.
Download: