Virtual Agents with Personality
If we really understand personality and interpersonal processes, couldn't we build characters who autonomously (by themselves) interact with other virtual and real others to produce outcomes that look like the behavior patterns of real people?  This is one of the questions that lead my collaborator, Stephen Read (in psychology at USC) and I,  into the realm of creating virtual agents with personality.  This question is interesting at a theoretical level, because we can't normally manipulate the parameters of mechanisms (e.g., those that are biologically inspired) and "see what happens" in complex social interactions. 

The question is also interesting at an applied level because it's relevant to creating realistic personalities for games (and other virtual environments), including "serious games" for training the work force or for educational or health purposes.  Imagine interacting with a personal coach, physician, virtual dates, virtual collaborators, and so forth and testing out and developing your interpersonal skills with agents with varying personalities (for example, suppose you could "tweak" different parameters akin to those associated with neuromodulators like serotonin).    You could test out your interpersonal skills on agents who respond quite differently -- with more or less in the way of self-regulatory skills --to the same situation. 

We are doing two types of work in this domain.  First, with a number of graduate students, including Aaron Brownstein, Brian Monroe, Yang Yu, and Anna Kostygina, we are interested --in general--- in connectionist models of social reasoning and behavior, and in developing a neural network model of personality.

Stephen Read and I are also working with Chi-Systems (Wayne Zachary), a Cognitive Architecture Firm in Philadelphia, on a project funded by the Air Force to create agents with realistic personality. 

 
Interactive Video to Reduce Sexual Risk-Taking                                                  back to top
Most HIV prevention interventions do not take into account the possibility that people are responding emotionally in sexual encounters.  Suppose instead you went on a virtual date in which you could make choices regarding what you would do next with a virtual other? Suppose if you were about to make a risky decision, you got a message crafted to reduce risky sex.  Would this message subsequently be activated when you were in the real-life risky context?  Would the intervention reduce the chances of engaging in risky sex? 

In our previous work, funded by the State of California University-wide AIDS research program, Stephen Read, Robert Appleby, and I  (and a number of my students) found that such an interactive video reduced risky sexual practices in a high-risk group of HIV negative men.  In our current funded work, we are trying to replicate this work and establish a template for developing interactive videos for health prevention.

Interactive Video Introductory Clip

Click here for standalone player
Download Windows Media Player Here