My research

At a high level, my research interests have evolved to be in the development and testing of technologies that enable non-technical users to perform complex technical tasks using computers. I am interested in developing artificial intelligence and machine learning techniques that enable domain experts to accomplish more with less effort. In addition, I am interested in the in vivo evaluation of those technologies through user studies.

Over the years I have found myself interested in a lot of research topics ranging from intelligence analysis, to human computer interaction, to combinatorics. The specific problems in those areas that piqued my interest are also consistent with my current research goals. Perusing the titles of my publications will give you an idea of some of my diverse research interests.

My dissertation addresses the design of compact representations and efficient algorithms to support the authoring of interactive virtual experiences such as computer games and interactive narratives. To realize these goals, my research draws on ideas from computer game design, machine learning, artificial intelligence, and social psychology. Interactive virtual experiences are computer-based experiences where a human participant (or multiple participants) interact with an environment defined by an author or designer. The ability for the participant to interact means that they have a direct effect on the environment and/or other participants in the experience. The specific experiences that I am studying are characterized by three components:

  1. A model for the experience. The model provides the rules that govern the interaction, the setting, and a context for the player to know what is acceptable. For example, the model could be a neighborhood environment. Houses and stores will be a part of it, players will know from the real world that conversations, dinner parties, mowing the lawn, etc. are acceptable and combat, violence, etc. are not. Further, the choice of rules will determine what is necessary for the player to succeed and how they can pursue success.
  2. A set of goals selected by the author. These goals are defined in terms of the possible interactions the player may experience with the environment or other characters. The choice of model provides a broad set of potential interactions. In many cases, especially when the experience is for training or education, only some subset of those interactions are desirable or consistent with the author's view of what should occur.
  3. An affordance for the player to exercise self-agency. Self-agency is the ability for the player to act in accordance with their own goals for the experience and be free from unnecessary or unwanted constraints.
Given the space of all possible interactions a player can have with a computer, the choice of model provides a restriction bias on the set of possible interactions. This is depicted as a rectangle in the image below, where each point in that rectangle represents an interaction. The specification of the author's goals provides a preference bias over the space of interactions consistent with the model. This is depicted in the image below by the green squares inside the rectangle. The player's self-agency affords goal-directed exploration of the space. This is depicted in the image below by the red squares inside the rectangle that are superimposed over the portions representing the author's goals. In the event that the player's goals do not align well with the author's goals, threats exist. A visual representation of threats is depicted in in the image below as the portion of the player's goals that are disjoint from the author's. The degree of these threats is motivation for what is referred to as Drama Management.


Motivation for drama management

A drama manager (DM) is an agent of the author that is tasked with shaping the player's experience. The image below depicts a schematic of the types of interactive virtual experiences I study. The author, as part of defining the model, creates an environment that the player interacts with and creates a set of non-player characters (NPCs) that populate and interact with the environment as well. The author also creates the DM to directly control or influence the NPCs as well as the environment. All of the DM's actions are in an attempt to get the player to pursue the goals they have that are consistent with the author's goals. Note, as depicted in in the image below, the DM can not directly influence the player, it can only make changes to the environment in an attempt to influence the player.

Schematic depiction of an interactive virtual experience

To accomplish my research goals, I draw upon ideas from various fields: the authoring process and evaluation platform I am developing are inspired by ideas from computer game design; the DM's decision making process is an algorithmic process which draws on ideas from artificial intelligence and machine learning; and the actions the DM can take are designed using theories of influence and persuasion from social psychology.

If you are interested in some more detail, please feel free to contact me.

Last update 1 - 10 - 2009