Modeling friendship formation, measuring peer effect and optimizing class assignment

Loading...
Thumbnail Image

Date

2024-03

Journal Title

Journal ISSN

Volume Title

Publisher

Research Projects

Organizational Units

Journal Issue

Abstract

Our research leverages neural network method to approximate the functional relationship between students' own characteristics and their friendship preferences, providing valuable insights into the intricate processes of how students form connections. This approach allows for an examination of friendship dynamics beyond what traditional linear models can offer. Secondly, replacing the linear-in-means assumption with a more reasonable friendship weighted assumption, we conduct a comprehensive analysis of heterogeneous peer effects based on students' characteristics. This allows us to discern the varying impacts of peer influence across different demographic and behavioral dimensions, contributing to a deeper understanding of peer effect distribution. Lastly, through counterfactual analysis, we go beyond describing observed patterns and actively identify optimal class assignment strategies that maximize students' cognitive scores. This practical dimension of our research holds implications for educational policies and practices, offering actionable insights for educators and policymakers seeking to enhance student outcomes through thoughtful class assignments.

Description

Poster Division: Social and Behavioral Sciences: 1st Place (The Ohio State University Edward F. Hayes Advanced Research Forum)

Keywords

Peer effect, class assignment

Citation