Promoting Pro-Social Behavior with End-to-End Data Science
user-5f03ee444c775ed682ef5240(2020)
摘要
Online platforms provide unprecedented opportunities to nudge pro-social behaviors: They track and host fine-granularity data generated by real-world users, which is a gold mine to understanding and modeling user behaviors. These interactive interfaces and rich functionalities provide excellent flexibility in implementing and delivering the nudge to the users. How shall we unleash the full potential of these platforms to nudge pro-social behaviors? In this dissertation, we propose an end-to-end data science pipeline that consists of three closely coupled stages: We first analyze user-behaviors with empirical data to discover potential nudges. We then develop recommender systems that maximize the effectiveness of the nudge with personalization. Finally, we implement the nudge in its original context and evaluate the nudge with randomized field experiments. Each stage of the pipeline calls for joint efforts from multiple disciplines, especially causal inference and machine learning. Moreover, the pipeline provides great flexibility for researchers to initiate their research, and make use of the latest development in causal inference and machine learning. We present three empirical studies conducted in distinct application contexts: an open-source software platform, an online microlending website, and a ride-sharing application. While they each start at a different stage along the pipeline, collectively, however, they demonstrate the effectiveness and flexibility of the proposed end-to-end pipeline in promoting pro-social behaviors.
更多查看译文
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络