Updated: 3/19/2006
It is claimed by some that evolution by natural selection is a "fact" instead of just a "theory". While it may be a very good theory, it has not reached fact status for the simple reason that it has not been fully observed. Observation of a claimed process actually taking place from start to finish should be the final lid on any theory to make into "fact". This is sometimes called "repeatability".
Evolution, however, has only given us slim slices of observation. We have observed minor changes in the field, such as bird beak color and size changes; or computer- simulated bigger changes under pristine, controlled software conditions. But nobody has demonstrated eyes and brains evolving from mud or organic soup in front of observers and cameras. True, it would probably take billions of years to finish such demonstration, but "time ate my homework" is not an acceptable excuse. If it cannot be fully demonstrated then it cannot be fully demonstrated. Apologies won't change that. The universe can be a real bummer at times and does not always want to make science easy.
The software business has taught me that there is no substitute for pressing the "run" button and watching the whole thing being used in actual production. Something that looks promising on paper, on a small scale, or in isolation can figuratively crash and burn in production. A portion may work well by itself in tests, but end up "not playing well with others". In software we have "unit tests" that test individual units of software. If the units pass their unit tests, then "integration testing" is the next step. Integration testing is to see if the units work well together. Passing unit tests is no guarantee that they will work right when finally put together.
Sometimes integration-testing exposes problems that require overhauling large parts of the design. It is not always easy to anticipate large-scale problems when designing systems on paper or building individual units. Often reality exposes misguided thinking. It is not always easy for the human mind to think of everything that can go wrong. Things can differ from expectations.
Pressing the "run" button is in some ways like a mini Judgment Day for software developers. Your work gets evaluated in ways you may not have anticipated and bunches of trees end up looking different when viewed as a forest or through the eyes of a different person. The reality check can sometimes be pretty jarring as months of effort finally sets foot onto the real world. We can relate to Dr. Frankenstein pulling the "awake" lever.
After integration testing, a test with actual users is arranged. "Users" are the people who actually would be using the software that is being built. Often they find issues that technicians and designers did not anticipate because technicians didn't understand what the users really wanted or didn't understand needs that the users had difficulty articulating.
The near-fatal trip of Apollo 13th was also a problem of integration rather than any one problem. It was the accumulative affects of short-sighted sensor design, technician inattention to power-usage data from an oxygen-flushing process, a minor shop-floor accident, and other minor slips. Any one of these small lapses by themselves would probably not doom Apollo 13. If the flush technician spotted the odd power-usage data in time, then the poor sensor design would not have mattered, for example. Similarly, if the sensor worked well, then there would not be any problem for the flush technician to spot. However, in conjunction all these mini-errors summed up to put the process over the threshold, and the tank burst in space.
In terms of evolution and observation it boils down to how the pieces of evidence are summed to make the final conclusion. On paper they do indeed seem to add up, and probably give the general right answer. However, we should keep a healthy skepticism. Truckloads of circumstantial and incremental evidence are good to have, but they are not enough to close the books on the case and lock it shut. There is just no substitute for full observation.
So, do scientists have a way to demonstrate live evolution at large scale without waiting a billion years? I believe the closest approach is computer-simulated evolution. Genetic algorithms that simulate the evolution process have indeed created some impressive results, solving some puzzles that are even difficult for people solve.
However, such simulations are still simulations. They may indeed demonstrate that the technique of breeding for winners can work as a principle under the right circumstances. However, that is still not the same as working in the wild; i.e., the "real world". Evolution simulations create artificially "clean" environments to simplify the process. They remove a lot of variables and complexities that real life has to face. This is to both speed up the process and because computers could not handle the complexity of full-out reality.
For example, the contestants of the simulated genetic races usually don't have to deal with bad weather, infectious diseases, and mating rituals. The genetic algorithm programmers omit a lot of this stuff so that the simulated organisms and selection mechanism can focus on the target goal and only the target goal. In many ways the programmers are playing God and cheating to make the contest simpler. The final result is thus actually a hybrid of evolution and Intelligent Design. We know that the simulated riders can ride using big training wheels, but that is not good enough to answer the final question of doing it in the real world.
All these other factors found in nature but clipped out to simplify the simulations perhaps may accumulate in a way similar to how the Apollo 13 errors accumulated to overwhelm the evolutionary process and make it not work. What may work with 5 factors to juggle may not necessarily work with 50 or 500. I doubt that is the case, but we must at least keep it in mind. "Probably" is not "fact".
The monster of intertwined complexity lurks in our software, our machines such as space capsules and nuclear power plants, and our world economy. It has shown to bite when we least expect it. Even the best nerds in the world have trouble mastering this monster. Otherwise, nuclear power plants would never melt down, we would know for sure what caused the Great Depression, prevented both space shuttle accidents, and Apollo 13 may have landed safely on the moon.
Until we can press the evolutionary "run" button and watch brains and eyes emerge from sludge, and with all the messy factors in place, not just those that we like; it stays a theory and not a fact. Evolution has been tested on a small scale and in simplified simulations, but it has not been battle-tested on the real front lines in front of cameras and observers with clipboards. Both history and Marvin Gaye with Tammi Terrell have taught us that there is "nothing like the real thing, baby".