Design and evaluation of a virtual environment infrastructure to support experiments in social behavior.

摘要

Virtual worlds provide useful platforms for social behavioral research, but impose stringent limitations on the rules of engagement, responsiveness, and data collection, along with other resource restrictions. The major challenge from a computer science standpoint in developing group behavior applications for such environments is accommodating the resource restrictions imposed on the developer by the virtual world implementation. The application design and development are strongly influenced by these restrictions, and turn them into a classic resource-limited computation problem with many novel properties.;The particular unique factors include a scripting language lacking many common data structures, limited runtime memory resources, built-in constraints designed to inhibit potentially dangerous user-scripted activities, limited network and communication protocols, and deliberately degraded speed and performance. The topology of the interaction among the controlling scripts is also limited in a unique fashion by the design of the virtual world available to us.;Our main contribution has been to frame a software environment to support the needs of the required social behavior experiments while adhering to these sometimes extreme limitations. The experiments that are supported by our virtual environment recreate prior studies done in clean, stylized environments that provided only minimally realistic social interaction. Our work satisfies the need to make the environment experienced by the subjects more realistic by adding the complexity of the virtual world and engaging the participants in more life-like decision-making. The analytic context that we have provided differs from previous work in online social networks and virtual worlds, in that we are able to conduct carefully controlled experiments, where specific parameters and rules of interaction between participants and the environment can be selected on a run-by-run basis.