Research

<-->

I became interested in research as an undergraduate when my advisor challenged me to use knowledge from both my communication sciences degree and my computer science degree to solve a single problem. My current research goal is to build a natural language generation system that uses the individual personality and agenda of a character to shape that character's communication and interaction with others. It is my belief that interactions are strengthened when we can build characters in games that take advantage of the human tendency to anthropomorphize virtual entities. I plan to extend the current state of the art for natural language generation in interactive games by infusing characters with personality and having the character filter a summary of game events based on a personal agenda. To make the character's summary more believable, I would like to impose an "incomplete knowledge" constraint; I will not allow any specific character to have global knowledge of the complete state of the game. This will create a subjective retelling of events that occurred in the human player's absence, colored by the character's experiences and agenda.

My interest in characters goes well beyond games. Game domains represent much of what makes the real world difficult: they involve human beings interacting with an environment and with each other; and to be successful, agents must integrate a broad range of capabilities, such as adaptation, reasoning, real-time reactivity, and especially communication. Games are an ideal domain in which to train or teach people without the consequence that may occur in real-life training. It is for these "simulation for training" experiences that my work designing realistic, human-like communication for computers will have its greatest impact.

My work on dialogue for belieavable characters is part of a number of research organization at in Georgia Tech's School of Interactive Computing. Here is a partial list of my affiliations:


Georgia Institute of Technology