Files in this item



application/pdfDEV-DISSERTATION-2021.pdf (3MB)
(no description provided)PDF


Title:Modeling the successes and failures of content-based platforms
Author(s):Dev, Himel
Director of Research:Sundaram, Hari
Doctoral Committee Chair(s):Sundaram, Hari
Doctoral Committee Member(s):Karahalios, Karrie; Zhai, ChengXiang; Shi, Xiaolin
Department / Program:Computer Science
Discipline:Computer Science
Degree Granting Institution:University of Illinois at Urbana-Champaign
Subject(s):content-based platform
knowledge market
platform success
platform failure
platform sustainability
content generation
voter biases
community voting norms
user roles
Abstract:Online platforms, such as Quora, Reddit, and Stack Exchange, provide substantial value to society through their original content. Content from these platforms informs many spheres of life—software development, finance, and academic research, among many others. Motivated by their content's powerful applications, we refer to these platforms as content-based platforms and study their successes and failures. The most common avenue of studying online platforms' successes and failures is to examine user growth. However, growth can be misleading. While many platforms initially attract a massive user base, a large fraction later exhibit post-growth failures. For example, despite their enormous growth, content-based platforms like Stack Exchange and Reddit have struggled with retaining users and generating high-quality content. Motivated by these post-growth failures, we ask: when are content-based platforms sustainable? This thesis aims to develop explanatory models that can shed light on the long-term successes and failures of content-based platforms. To this end, we conduct a series of large-scale empirical studies by developing explanatory and causal models. In the first study, we analyze the community question answering websites in Stack Exchange through the economic lens of a "market". We discover a curious phenomenon: in many Stack Exchange sites, platform success measures, such as the percentage of the answered questions, decline with an increase in the number of users. In the second study, we identify the causal factors that contribute to this decline. Specifically, we show that impression signals such as contributing user's reputation, aggregate vote thus far, and position of content significantly affect the votes on content in Stack Exchange sites. These unintended effects are known as voter biases, which in turn affect the future participation of users. In the third study, we develop a methodology for reasoning about alternative voting norms, specifically how they impact user retention. We show that if the Stack Exchange community members had voted based upon content-based criteria, such as length, readability, objectivity, and polarity, the platform would have attained higher user retention. In the fourth study, we examine the effect of user roles on the health of content-based platforms. We reveal that the composition of Stack Exchange communities (based on user roles) varies across topical categories. Further, these communities exhibit statistically significant differences in health metrics. Altogether, this thesis offers some fresh insights into understanding the successes and failures of content-based platforms.
Issue Date:2021-04-19
Rights Information:Copyright 2021 Himel Dev
Date Available in IDEALS:2021-09-17
Date Deposited:2021-05

This item appears in the following Collection(s)

Item Statistics