The stabilization of environments

摘要

In planning and activity research there are two common approaches to matching agents with environments. Either the agent is designed with a specific environment in mind, or it is provided with learning capabilities so that it can adapt to the environment it is placed in. In this paper we look at a third and underexploited alternative: designing agents which adapt their environments to suit themselves. We call this stabilization, and we present a taxonomy of types of stability that human beings typically both rely on and enforce. We also taxonomize the ways in which stabilization behaviors can be cued and learned. We illustrate these ideas with a program calledFixPoint, which improves its performance over time by stabilizing its environment.