Understanding human behavior using interactions with online content
Abstract/Contents
- Abstract
- People often interact with one another on online applications through liking, viewing, and sharing each other's content. Instead of these applications being comprised only of a network of individuals, the content itself often plays a primary role. One can view such applications in terms of a content interaction graph: a bipartite graph with users on one side, content on the other, and edges representing interactions with content connecting the two. These content interactions reflect user intentions, behavioral patterns, and biases. Understanding these behaviors enables improved content recommendation and user interface flows, which in turn provides increased utility to users of content-driven online applications. In this thesis, we study in various ways how the content interaction graph can be used to better understand the behavioral mechanisms through which humans interact with content. First, we examine individual user intentions by studying users' purchasing intent, i.e. inclination to make a monetary purchase, on Pinterest, a content discovery application. We show how user activity information, the way users interact with content, and content itself can be used as signals of rising purchasing intent over time. We next focus on content saving interactions, by studying the way users save content into collections over time. We characterize the growth of these individual collections, and show how signals obtained from modeling the inter-event times of saving interactions can be used to predict collection size and lifespan. Finally, we broaden the scope of "users" in content interaction graphs to media outlets (e.g. CNN) to show how such a graph can uncover media bias in the coverage of Obama's presidency. We discuss NIFTY, a large-scale system to efficiently identify and track "memes", or quoted content, that mutate over time. We then show how clusters of quoted content can be encoded in a low-rank space that reveals dimensions of media bias in an unsupervised manner. Put together, this work demonstrates how content interaction graph analysis is a powerful tool that can, in the future, be used to infer an even broader range of user behaviors on online applications.
Description
| Type of resource | text |
|---|---|
| Form | electronic; electronic resource; remote |
| Extent | 1 online resource. |
| Publication date | 2018 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Associated with | Lo, Caroline |
|---|---|
| Associated with | Stanford University, Computer Science Department. |
| Primary advisor | Leskovec, Jurij |
| Thesis advisor | Leskovec, Jurij |
| Thesis advisor | Garcia-Molina, Hector |
| Thesis advisor | Jurafsky, Dan, 1962- |
| Advisor | Garcia-Molina, Hector |
| Advisor | Jurafsky, Dan, 1962- |
Subjects
| Genre | Theses |
|---|
Bibliographic information
| Statement of responsibility | Caroline Lo. |
|---|---|
| Note | Submitted to the Department of Computer Science. |
| Thesis | Thesis (Ph.D.)--Stanford University, 2018. |
| Location | electronic resource |
Access conditions
- Copyright
- © 2018 by Caroline Lo
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...