Science Cannot Validate Itself
Science is an unbiased, objective, disciplined, cooperative method for progressively uncovering truth about the natural world. That’s the way most of us were taught to think about it in school. Further reflection, however, produces a host of questions rarely discussed in science class. How does science differ from other unbiased, objective, disciplined, cooperative methods of inquiry? What is special about scientific logic? To what does science refer? How much impact does our humanness and our relationships have on scientific theories? What is the scientific method? How is science to be distinguished from pseudoscience? Are all branches of science worthy of the same respect? What constitutes a scientific explanation? If our best theories are only tentative, how can we ever know when we have a grasp on reality that is unlikely to be overturned or subsumed under a greater theory? These and many other questions can keep philosophers of science in the Humanities departments busy for years (but with less grant money). Working scientists don’t often pay them much attention. Maybe they should.
Nature1 printed a rare excursion into philosophy of science2 that cast severe doubt on the ability of science to ever grasp reality with sufficient confidence to say we have “arrived” at understanding of the cosmos. P.-M Binder, an astronomer at the University of Hawaii in Hilo, explored the reasonings of David Wolpert, known for his work on the “No Free Lunch” theorems.3 He sought to explore the nature and limits of scientific reasoning. Wolpert demonstrated in a recent paper4 that “the entire physical Universe cannot be fully understood by any single inference system that exists within it” (Binder’s words). If that sounds like something Turing or Gödel would say, it is.
Wolpert is not the first to demonstrate fundamental limits on human knowledge. Heisenberg’s Uncertainty Principle is a famous example. Gödel’s Incompleteness Theorem is another: it placed fundamental limits on the ability of mathematical theories to validate themselves. Wolpert follows in this tradition with “impossibility results.” He proved with mathematics and logic that in the Universe of sequences of events that follow natural laws, no two strong inference machines can be strongly inferred from each other. His conclusions are independent of any particular natural laws employed in the inference.
This means that science can never know everything: just almost everything in the best case. When you “know” one inference well, there will always be at least one other category of inference that will be unclear or ambiguous. Example: the equations of chaos theory can perform pretty well in predicting outcomes of seemingly disorganized systems that have a “strong attractor,” at least up to an acceptable level of accuracy. The catch is: the method cannot validate the equations themselves. What Wolpert has done, in his own words, is demonstrate “impossibility results” in scientific logic that “can be viewed as a non-quantum-mechanical ‘uncertainty principle.’”
In short, science cannot validate itself. Science will never produce a theory of everything. Gone are the optimistic 18th-century traditions of Laplace that, given knowledge of each particle’s position and momentum, future outcomes could be predicted with any desired degree of certainty. The Uncertainty Principle, generalized into scientific logic by Wolpert, has shown that the more precise an observer measures one quantity (or inference), the more uncertain becomes the other. Gone also are claims that given a long enough lever and a place to stand, one could move the world. That standing place will always be wobbly.
1. P.-M Binder, “Philosophy of science: Theories of almost everything,” Nature 455, 884-885 (16 October 2008) | doi:10.1038/455884a.
2. Two articles in The Scientist this month affirm that philosophy of science is neglected in science education these days: one by James Williams on “What Makes Science ‘Science’?” and a follow-up by Richard Gallagher on “Why the Philosophy of Science Matters.” Both articles, unfortunately, appear to espouse a narrow view that resembles logical positivism. This view would be considered indefensible by many philosophers today after the Kuhnian Revolution of the 1960s and the Science Wars of the 1990s. Both also arrogated objectivity to establishment scientists while denigrating creationists and others as ideologues. One respondent caught Summers in name calling. Summers backpedaled somewhat, acknowledging his own dogmatism and the fallibility of science.
3. For background on the No Free Lunch theorems, see William Dembski’s book No Free Lunch, Rowman and Littlefield, 2002, esp. section 4.6.
4. David H. Wolpert, “Physical limits of inference,” Physica D 237, 1257�1281 (2008), doi:10.1016/j.physd.2008.03.040.
“Science is truth,” chants Finagle’s Creed; “Do not be misled by facts!” The limitations of scientific inference explored by Wolpert must hit thinking scientists like a rude awakening. It’s like dreaming of climbing a mountain only to find oneself going up a down escalator. The Truth about the Universe will forever remain beyond the reach of science.
Binder ended on a confident note that science might still be converging on a close approximation of reality. Oddly, he ended by showing that two subjects in fundamental physics are beset with shortcomings: the standard model of particle physics, and the so-far intractable problem of uniting quantum mechanics with gravity. But then he said optimistically, in conclusion, “It is possible, though, that these various theories, along with all that we have learned in physics and other scientific disciplines, will yet merge into the best science can do: a theory of almost everything.”
Almost is not good enough. There will always be something else you cannot know. Like Ken Ham quips: if you can’t know what you don’t know, you can’t know what you do know; and if you can’t know what you do know, you might know very little. To which we add: how could you ever know whether the most important puzzle piece lies outside your world view, in the inference machine that cannot be inferred from within your system? Maybe, for instance, the most important piece lies in theology.