Quantifying Context and its Effects in Large Natural Datasets
- Author(s): Vinson, David W.
- Advisor(s): Matlock, Teenie
- et al.
Intended outcomes such as expressing ideas in ways that can be understood, with the tools we know how to use, constrain the dynamics of the actions we use to ensure their success. The success of our actions can be measured by estimating the amount of information they transmit. This provides a window into the dynamics of cognition and how they are influenced by the surrounding context: Including past, present and future actions, cognitive states, social pressures and the tools we use to generate them —such as language. This dissertation is a series of studies on how context influences intended outcomes including current decisions, action dynamics and the amount of information that can be transmitted successfully. The focus is primarily on information and how it influences —and is influenced by —behavior.
The study of information cuts across disciplines. As a result, this dissertation consists of a set of interdisciplinary collaborations that quantify and explore large natural datasets. In collaboration with various coauthors from linguistics, cognitive science and applied mathematics, I present five projects that address theoretical and methodological approaches toward understanding the effects of context and what they might say about the success of our behaviors in bringing about intended effects. The studies here are presented chronologically in an effort to elucidate the process of science as it occurs naturally.
First, I present a study exploring how the Information-Theoretic structure of a message is influenced by its intended valence (Vinson & Dale, 2014b) followed by how it is influenced by social network structures (Vinson & Dale, 2016). I then report on the development of a new analysis tool that affords quantifying large text data efficiently (Vinson, Davis, Sindi, & Dale, 2016). This is followed by a study on how the dynamics of action —captured via the process of typing a message —are not encapsulated, but dynamically adjust to the statistics of the language they are used to produce. (Vinson, Dale, Shih, & Spivey, submitted). Finally, I present a study exploring how previous online business review ratings influence current ratings (Vinson, Dale, & Jones, in revision).
This dissertation, Quantifying Context and its Effects in Large Natural Datasets, is submitted by David W. Vinson in partial fulfillment of the degree Doctor of Philosophy in Cognitive and Information Sciences at the University of California, Merced, under the guidance of dissertation committee members Rick Dale, Teenie Matlock and Jeff Yoshimi.