Bayesian models of design based on intuition

Abstract

Most computer system designers use a great deal of intuition in the design process. Intuition is often used to handle uncertainty in design parameters. Since uncertainty seems to be intrinsic to most design problems it follows that designers will continue to rely on intuition or “sound engineering judgement”. This paper attempts to use Bayesian Decision Theory to explore the possibility of setting up a structure and theory for making design decisions in the computer system design environment while explicitly taking the intuitive nature of many design decisions into account. We shall focus attention on a particular problem in distributed data base design in which the designer must use his intuition to estimate the load on the system which he is designing. Similar Bayesian approaches could be used in other design problems.