Many online human activities leave digital traces that are recorded in natural-language format.  The exploitation of this unprecedented resource under a computational framework can bring a phase transition in our understanding of human social behavior and shape the future of social media systems.  This thesis describes a computational approach to an intriguing aspect of conversational behavior, linguistic style coordination: conversational participants tend to immediately and unconsciously adapt to each other's language styles; a speaker will even adjust the number of articles and other function words in their next utterance in response to the number in their partner's immediately preceding utterance.

We develop a probabilistic framework that can model and quantify this phenomenon and  apply it to four different conversational settings:  Twitter conversations, discussions between Wikipedia editors, oral arguments in the U.S. Supreme Court and movie script dialogs.   We are able to examine linguistic style coordination in large scale settings for the first time, and by doing so, to reveal important  properties that were never observed in previous studies, which were conducted mainly in small and controlled environments. 

The resulting new understanding of conversational behavior bears practical importance in the task of uncovering key properties of social relations.  We present an analysis framework based on linguistic coordination that can be used to shed light on power relationships and that works consistently across multiple types of power --- including a more "static" form of power based on status differences, and a more "situational" form of power in which one individual experiences a type of dependence on another.

Finally, our computational framework can be employed to gain insight into the causal mechanism behind coordination behavior. It was previously hypothesized in the psycho-linguistic  literature that coordination has arisen as a way to achieve social goals, such as gaining approval or emphasizing difference in status.  But has the adaptation mechanism become so deeply embedded in the language-generation process as to become a reflex?  We argue that fictional dialogs offer a way to study this question, since authors create the conversations but don't receive the social benefits (rather, the imagined characters do).  Indeed, we find significant coordination across many families of function words in our large movie-script corpus.  We thereby provide evidence that language coordination is so implanted within our conception of conversational behavior that, even if such coordination is socially motivated, it is exhibited even when the person generating the language in question is not receiving any of the presumed social advantages.

@phdthesis{Danescu-Niculescu-Mizil:thesis2012,

  author={Cristian Danescu-Niculescu-Mizil},

  title={A Computational Approach to Linguistic Style Coordination},

  school={Cornell University},

  year={2012}

}