Problem solving in a computational society

摘要

We propose a new framework to study problem solving in a computational society. Such a society consists of computational agents able to use information about the direct and indirect computational experiences of other agents. Our history-based computing framework generalizes existing computational frameworks that might be used in a computational agent society, like relativized or oracle computing and knowledge compilation (KC). We present a preliminary classification of problems using this framework and give some examples showing that direct compilation of computational experiences of other agents can be significantly enhanced by adding meta-information about these experiences. Finally, we discuss some relationships with relativized computing and KC, showing that our framework is able to refine both.