Can we do away with judgment/intuition in systems design?

The judgement inherent in this balancing of programs and systems can no longer be intuitive or rely on past experience alone. The range of choice is too broad; the number and type of alternatives too great.

 

In the selection of weapon systems, in the design of forces, and in determination of the level of the national defense effort, therefore, we are making greater use of a technique called systems analysis. Perhaps it is best described as “quantitative common sense.”

 

Systems analysis takes a complex problem and sorts out the tangle of factors. It aims to assist the decision-maker by furnishing him with quantitative estimates of the effectiveness and costs of each of the alternative courses which he could choose. Confronting a multiplicity of options we have turned to analytical techniques to assist us in our choice…

 

I am sure that no significant military problem will ever be wholly susceptible to purely quantitative analysis. But every piece of the total problem that can be quantitatively analyzed removed one more piece of uncertainty from our process of making a choice. There are many factors which cannot be adequately quantified and which therefore must be supplemented with judgment seasoned by experience. Furthermore, experience is necessary to determine the relevant questions with which to proceed with any analysis.

 

I would not, if I could, attempt to substitute analytical techniques for judgment based upon experience.

Robert McNamara writing in the Civil Service Journal, Vol. 4, No. 4, April-June 1964, pp. 1-5. Note the reductionist viewpoint McNamara holds. Problems of weapons choice — which he admits is too big to be decided purely through analysis — can be disaggregated into small problems which lend themselves to quantitative optimizations. Then the answer is supposedly to put the puzzle back together. Successions of suboptimizations get you to an optimal answer.

Of course, in reality there are nonlinearities in system interactions such that the whole isn’t simply the sum of its parts. Further confounding our problem is that alternatives have attributes which are incommensurable, or cannot be put into common units for comparison.