Computers and Scientific Proofs

Today’s movie directors use computer technology to take us to wondrous places.  We can see J. R. R. Tolkien’s Middle Earth, C.S. Lewis’ Narnia and James Cameron’s Pandora come to life in vivid detail with the aid of high speed computers “painting” each pixel of the screen with exactly the right color to make these worlds of fantasy as real as scenes of our daily lives. We willingly suspend our disbelief to enjoy our “journeys” to these worlds of fantasy.

We’ve all seen the results of computers being used to create a fantasy world. It should therefore be obvious that computer output can’t be used to prove, or disprove, any scientific theory. Those who claim that computer outputs provide any kind of evidence in the Global Warming debate are in error.

The is nothing in the architecture and technology of computers which guarantees the accuracy of the results. Programmers, or more specifically, the people who pay the bills, specify the degree of reality desired when they determine what kind of program is wanted. When fantasy is desired, that can be accomplished. Various levels of accuracy and, if you will, truth can be sought, even if they are not attained.

At the basic level, computer programs only do a few things.  They gather information and can do calculations. Then, based on the rules they are given, they do one thing or another, and present the results. An example would be the following program which does arithmetic correctly, unless it is April Fool’s Day.

     Read numbers A and B

     If April-Fools-Day and

           A = 2 and B = 2

       then C = 5

       else C = A + B

     End-if

     Print A + B = C

On April 1, this will print 2 + 2 = 5. On any other day, it will print 2 + 2 = 4.

The rules given to the computer in a program determine the accuracy of the results. Computers are especially effective in record-keeping. In many ways they resemble accurate clerks who do precisely what they are told, and do it very quickly.  Any industry with a heavily clerical work environment can use them effectively.  Sometimes a computer error is a result of data entry. The wrong amount is entered. The rules coded into the computer can also produce the wrong results.

At best, computer programs can reflect the current state of knowledge of a situation. Sometimes, even when the understanding is accurate, it can change under you. A current case is the definition of marriage.  For years, programmers could validate a computer-recorded marriage by confirming that one party was male and the other was female.  In some jurisdictions, this is no longer true.  Programs have to be changed to reflect the new facts.

Scientific models are subject to these same problems.  In a ideal world, we would know all the variables affecting a situation and be able to program the cases correctly.  But in science, there are always new findings and the discovery of new rules.  The rules used in today’s computer model can only reflect the current understanding of the problem.  Tomorrow’s knowledge will mean that today’s program has to be changed.

Computers are used in the design of everything from airplanes to buildings. Doesn’t this show that we can trust their results? No, it doesn’t. Again the computer’s results are only as good as the input and the rules. Computer modeling can save the time and cost of wind-tunnel testing of an airplane design. Changes to the proposed design can be made and tested easily. But the results are only as good as the input and the assumptions about how things work. Thankfully, the “final” designs are subjected to wind-tunnel testing and courageous pilots test the plane in flight before the plane is certified.

The value of computers in engineering and science is the ability to rule out certain alternatives and suggest a possible right answer. This is useful. But, to go beyond that and claim that computer models can, in any way, prove a scientific matter, is to give the computer an authority it doesn’t and can’t have.

Proofs of scientific theories need repeatable measurements and tests, computer models are a useful form of inquiry, but they are not proofs.