Machine Learning for Collective Behavior

这项研究将开发和应用新的机器学习方法来模拟集体问题解决背景下的个人行为。在早期的人类实验研究中，实验任务是仅从局部信息解决网络上具有挑战性的分布式计算问题，例如图着色。新工作的目标是开发机器学习方法，其输出可以准确地从个体模型中重建和预测集体行为，并可以揭示相关问题，如人类群体中策略的经验多样性，及其对有效集体行为的重要性。该项目以新颖的方式为几个不同的研究社区做出贡献，包括机器学习，社会学，经济学和相关领域。它以两种方式将研究和教育结合起来：一是为研究生和本科生提供参与尖端研究的体验，二是为一门名为“网络生活”的课程提供新课程。

This research will develop and apply new machine learning methods to model individual behavior in the context of collective problem-solving. The experimental task in earlier empirical research with human subjects was to solve challenging distributed computational problems on networks, such as graph coloring, from only local information. The goal of the new work is to develop machine learning methods whose outputs can accurately reconstruct and predict collective behavior from individual models, and can shed light on related questions such as the empirical diversity of strategies within a human population, and its importance for effective collective behavior.This project contributes in novel ways to several distinct research communities, including machine learning, sociology, economics, and related fields. It integrates research and education in two ways: by giving both graduate students and undergraduates the experience of participating in a cutting-edge research study, and by providing new curriculum for a course entitled "Networked Life."