Food webs are abstract models that represent who eats whom relationships in ecosystems. Classical food web representations do not typically include phenotypic plasticity, in which one species responds to changes in density of other species by modifying traits such as behavior and morphology. Such changes, which are presumably adaptive, will affect the magnitude of both direct and indirect effects on species fitness. Empirical evidence suggests that phenotypic plasticity is likely to have large impacts on the structure and dynamics of ecological communities. Whereas theoretical studies support this, there is much that we do not understand that may require new theoretical approaches. We have constructed a computational system, Digital Organisms in a Virtual Ecosystem (DOVE), to address this problem. Its features include an individual-based approach, in which a type of genetic algorithm is used to evolve animal behavior in a dynamic environment. Here we present an overview of the ecological problems motivating the creation of DOVE and its basic structure and operation. We also discuss the kinds of decisions and tradeoffs that were considered to make DOVE as simple as possible but still rich enough to allow us to address our fundamental questions. We then use DOVE to examine optimal foraging strategies of prey in the presence of fluctuating predator risk, and show that activity levels are highly dependent on competitor density in a manner that would be difficult or impossible to explore with traditional techniques. This, and other pilot studies of DOVE, suggests that it can be used to gain insight into the origin and consequences of phenotypic plasticity and other properties of ecological communities.