Intelligent environments are physical spaces that can sense and respond to the people and events taking place within them, providing opportunities for people to influence environmental factors that affect them, such as the lighting, temperature, décor or background music in the common areas of an office building. The designer of an environment that can be influenced by a group of collocated people rather than a single individual must decide how to accord influence among the individuals in the group. We have designed two multi-agent group preference arbitration schemes and tested them out in an intelligent environment, MUSICFX, which controls the selection of music played in a fitness center. One scheme seeks to maximize the average satisfaction of the inhabitants, the other seeks to maximize the equitable distribution of satisfaction among the inhabitants. We present the results of a series of experiments using real data collected from the deployed system, and discuss the ramifications of these two potentially conflicting goals.