Synthetic Social Relationships for Computational Entities

Humans and many other animals form long term social relationships with each other. These relationships confer a variety of benefits upon us, both as individuals and as groups. Computational systems that can form social relationships like those formed by animals could reap many of the benefits of sociality, both within their own groups and in their interactions with people. This dissertation explores two main questions: • What kinds of internal and external representations are necessary for computational entities to form social relationships like those formed by animals? • How can people participate in and direct the relationships of these entities? To explore these questions, I designed and implemented a system by which computational entities may form simple social relationships. In particular, these synthetic social relationships are modeled after the social behavior of the gray wolf (Canis lupus). The system comprises a novel combination of simple models of emotion, perception and learning in an emotional memorybased mechanism for social relationship formation. The system also includes supporting technologies through which people may participate in and direct the relationships. The system was presented as an interactive installation entitled AlphaWolf in the Emerging Technologies program at SIGGRAPH 2001. This installation featured a pack of six virtual wolves – three fully autonomous adults and three semi-autonomous pups whom people could direct by howling, growling, whining or barking into microphones. In addition to observing the interactions of several hundred SIGGRAPH participants, I performed two main evaluations of the AlphaWolf system – a 32-subject human user study and a set of simulations of resource exploitation among the virtual wolves. Results from these evaluations support the hypothesis that the AlphaWolf system enables the formation of social relationships among groups of computational entities and people, and that these relationships are beneficial to both the inter-machine interactions and the human-machine interactions in a variety of ways.