Files in this item

FilesDescriptionFormat

application/pdf

application/pdfJuan_Mancilla Caceres.pdf (2MB)
(no description provided)PDF

Description

Title:Social sensing games
Author(s):Mancilla Caceres, Juan
Director of Research:Amir, Eyal
Doctoral Committee Chair(s):Amir, Eyal
Doctoral Committee Member(s):Espelage, Dorothy L.; Girju, Roxana; Karahalios, Karrie G.; Lieberman, Henry
Department / Program:Computer Science
Discipline:Computer Science
Degree Granting Institution:University of Illinois at Urbana-Champaign
Degree:Ph.D.
Genre:Dissertation
Subject(s):Game-Based Methods
Social Network Analysis
Identification of Bullying and Cyberbullying
Evaluation of Commonsense Knowledge
Formalization of Social Sensing through Games
Global Inference from Pairwise Interactions
Visualizing Pairwise Interactions
Combining observational methods with lab-controlled experiments in a computational setting
Abstract:We introduce Social Sensing Games (SSGs), a new method for collecting data about social relationships, and present an algorithm that can be used to efficiently infer information from the output of the games. The main purpose of this new method is to use people's online interactions to learn about their offline behavior. Traditionally, scientists have studied social interactions through the use social networks obtained through carefully designed self-report surveys that impose limitations on the types of research questions that can be answered. Recently, thanks to the ever-increasing use of computers and mobile devices for managing social relationships, scientists look to use large amounts of data that is easily accessible. Unfortunately, this latter data lacks the experimental and theoretical validity of previous methods. SSGs address these concerns by providing an interface that can collect fine-grained data that is relevant to the research question at hand. The first contribution of this thesis is the formalization of Social Sensing Games in such a way that they combine the power of lab-controlled experiments, the detailed observations of observational studies, and the scalability and inferential power of computational methods. This new definition can be used by researchers to easily design SSGs specific to the problem they wish to address. The second contribution is most relevant to the field of social network analysis. We present an algorithm for analyzing the output of SSGs which main insight is that, in some cases, pairwise relationships are enough to infer global attributes of the nodes encoded in a social network and that such assumption may reduce the complexity of inference, help with the scarcity of data, and still maintain some of the context of the network. We show these contributions through two applications: The evaluation of commonsense knowledge, and the identification of classroom aggressors (or bullying). The second application being in itself an important contribution that provides new insights concerning the study of bullying and cyberbullying.
Issue Date:2014-05-30
URI:http://hdl.handle.net/2142/49616
Rights Information:Copyright 2014 Juan Fernando Mancilla Caceres
Date Available in IDEALS:2014-05-30
Date Deposited:2014-05


This item appears in the following Collection(s)

Item Statistics