AIC Seminar Series

Stemming from an initial realization that humans tend to personify complex automation regardless of whether it is given a "face" or an embodiment, we have had a growing interest in understanding the implications of patterns of social interaction and expectation (what we call "etiquette") for human-computer interaction design. This interest has led us to develop a computational model of a significant aspect of etiquette– politeness and its role in establishing, maintaining and revising social relationships. Our model, based on the socio-linguistic work of Brown and Levinson (1986), posits a culturally universal function for politeness– the management of "face threats" in interactions between intentional agents– and explains the role of politeness in power and familiarity relationships and in the conveying of imposition, urgency, indebtedness, etc. We have developed a computational implementation of this model and have demonstrated its ability to deliver culturally-specific perceptions of politeness in game and simulation settings. More recently, however, we have extended this model to include the impact of perceived politeness on decision making, attitudes and behaviors in response to directive speech acts. Through a series of human subjects studies, we have the beginnings of a model of how the politeness of a directive will be perceived in context by members of different cultures, and given a level of perceived politeness, what measurable effects it will have on variables such as trust, affect, perceived workload, memorability, reaction time, accuracy and overall compliance likelihood. There are a wide variety of applications for this modeling approach, some of which we are beginning to develop.

Dr. Christopher A. Miller is the Chief Scientist and a co-owner of Smart Information Flow Technologies (SIFT) based in Minneapolis, MN. He has more than 20 years experience in R&D for advanced human-automation interaction systems– first at Honeywell’s Technology Center and then at SIFT. With a degree in Cognitive Psychology from the University of Chicago and an ongoing interest in Artificial Intelligence, much of his work focuses on developing and applying computational models of human cognitive phenomena. Examples include a computational metric for information "fit" which has been used to dynamically adapt cockpit displays to a pilot’s ongoing task needs; an approach to human-machine interaction, called "Playbook(R)", which uses a shared hierarchical task model as an expandable lingua franca between a human supervisor and automation (such as Unmanned Air Vehicles) and provides the human with the ability to flexibly delegate authority to automation at either high or low levels of control; and a culturally universal model of perceived politeness, it’s role in managing social relationships and its impact on directive compliance behaviors.