Adaptive resource management (ARM) technology has emerged as a way to: (1) support applications? varying resource needs, and (2) provide reusable solutions to commonly occurring problems. Most existent works on ARM have focused on centralized approaches. However, centralized ARMs are not scalable, which may result in slow responses and wrong allocation decisions based on inaccurate system state information. In this paper, we show how to address these problems with a decentralized resource management approach. Based on social models, the approach appears to have the potential to provide prompt and accurate allocation of resources.