Philip Sargent
EDRC
Wednesday, September 10, 1997

Instead of endlessly trying to think up what we mean by design science, why don't we try to imagine what some typical individual's views might be ? This is easier because we can imagine people with limited viewpoints, and it is useful because we can imagine people who might give us money. Some of these viewpoints are self-contradictory, but that doesn't mean that real people don't believe in them.

I want Design Science to tell me how to design products better and faster, and more cheaply too. If some guy tells me that all I have to do is to split off a small team and set up a skunk works because that is a proven sucessful method then I'll spit in his eye. I want something "scientific" by which I mean some magic process which will improve my design shop without having to reorganise my organisation, fire anyone or make any painful changes. I don't care about fundamental basics, methodologies or all that stuff, I want something that works.

I want some set of administrative procedures that will tell me which of the researchers are making progress and which projects should be junked. I want something abstract that will tell me the fundamental worth of a line of research even if the guy doing it is a jerk and can't make it work, and I want something that will tell me if some bright people are stuck in a dead end and should be redeployed. I want some measures so that I can show that my laboratory is making progress. I don't want anything that means I have to spend a lot of effort analysing failed projects, I want to concentrate on positive, new advances. I want some objective and irrefutable method to justify the decisions I want to make.

The relationships between hypotheses, theories and experiments are in such epistimological chaos that the best one can hope for is a "scientific" attitude: where investigators are objective, systematic and rational but where the subject studied need not be any of these.

When I have developed Design Science I will know which techniques to use to design anything, how to generate variant designs and how to decide between them. I will know how to select people, how to organise them and which tools they will find useful. This knowledge will be expressed in a few simple statements that are equally applicable to mechanical, electronic, software, architectural and legal-system design.

A science of the design is hypothesized to be "a body of intellectually tough, analytic, partly formalizable, partly empirical, teachable doctrine about the design process". It includes utility theory, linear programming, dynamic programming, techniques for optimization, resource allocation, finding alternatives and finding appropriate representations.

I want something that I can teach in a course, something that is easy to examine and something that won't upset all the other professors. I want to present the whole design process to students with minimal experience of engineering, commerical organisations, economics or social psychology and I want to make it realistic and I have only 3 hours in the curriculum to do it. I want to be systematic and complete. Some memorable diagrams and a small number of key insights would help.

I want to understand what goes on in a designer's head. Design Science will describe what sequence of thoughts a designer has as he does something creative, how he paces himself and why some people are better at it than others. With Design Science I will be able to train engineers to be better designers using neat little thought-exercises, and I will be able to screen people for designing aptitude easily, quickly, reliably and quantitatively.

I want to understand what designers do when they work together. When I understand this I will know how an organisation produces good designs and which organisations cannot produce good designs. I will know which non-verbal signs designers use to express disapproval, how they conduct initiates into their craft, how they learn to trust each other, and what the inner meaning is of their forms of ritual combat.

Design science will give me a well-ordered set of techniques to embody in software that will enable me to completely automate any design task. Design science will be a taxonomy of data-structures, algorithms and interface specifications.

I want something means I have to spend less time in meetings and more time doing design. I want to get on with the job and spend less time doing things that are not important. I want to design really neat things. I don't want to be sent to any training course where the word "methodology" is used. I was trained as a <...> engineer, I want to spend my time doing <...> engineering.

Philip Sargent
EDRC, CMU