FreeStyle:The Structure of a Scientific Concept
From ResearchID.org, a nexus for researching Intelligent Design
The goal of science is to generate predictions about what the universe around us - what might be called "consensual reality" - will confront us with next. At some level, it is something we all do from an early age - that is how we build up our understanding of the universe. However, in recent years, the philosophy behind this process has become increasingly rigorous.
An unconfirmed statement about the objective universe that matches existing data is usually referred to as a conjecture.
Contents |
Usefulness criteria
There are some conjectures that just are not terribly useful. For example, the conjecture that there is another universe parallel to our own that we can in no way access is not useful to science. The conjecture that the universe was created five minutes ago but was designed to look like it was billions of years old is not useful to science. The conjecture that humans possess souls that are completely undetectable in any fashion is not useful to science.
When speaking in the context of a Popperian model of science, there are two different usefulness criteria: falsifiability and verifiability. If the truth of a conjecture can be confirmed, that conjecture is verifiable. If the truth of a conjecture can be refuted, that conjecture is falsifiable.
Examples:
- The conjecture "there exists a black horse" is verifiable - you just need to find one example of a black horse.
- The conjecture "all horses are black" is falsifiable - you just need to find one example of a white horse.
- The conjecture "the next horse I see will be black" is both verifiable and falsifiable - it will be confirmed or refuted the moment you set eyes on a horse.
Verifiable conjectures are generally of the form "there exists an object X in set S with property Y". Falsifiable conjectures are generally of the form "all objects in set S have the property Y". Conjectures that are both verifiable and falsifiable are generally of the form "object X has property Y".
Applying the basic rules of induction, we can see that the negation of a verifiable conjecture will be falsifiable, and vice versa. The negation of an unverifiable conjecture will, correspondingly, be unfalsifiable.
Conjectures which are both verifiable and falsifiable are referred to as predictions. The negation of a prediction will also be a prediction.
Note that there are in fact two ways in which a conjecture can be unfalsifiable. Conjectures of the form "there exists an object X with attribute Y" will always be completely unfalsifiable - that object could exist in an alternate universe for all we know. Conjectures of the form "there exists an object X in set S with attribute Y" will not necessarily be unfalsifiable - for example if the set S was the set of coins currently in your pocket and the attribute Y was being a pound. You can falsify that conjecture by exhaustively examining every member of the set S. This only works, however, if S is sufficiently small and every member of S can be examined. If this is not the case then the conjecture is effectively unfalsifiable.
Hypotheses
Conjectures that are only verifiable are not very useful. Consider the conjecture "there exists a teapot floating somewhere in the asteroid belt". It is easily verifiable - you just need to find the teapot - but it is effectively unfalsifiable - you cannot examine every single object in the asteroid belt to check if it is a teapot. Unfalsifiable conjectures are impossible to eliminate, so you will always have millions of them floating around (literally, in the case of the teapot) and there is no way to tell which are "best" unless/until they have been confirmed. Hence, science generally does not get involved with these - it is far more interested in falsifiable conjectures.
Note that verifiable conjectures, once verified, become data, and are part of the initial test used to confirm that a statement is a valid conjecture.
A conjecture that is falsifiable is referred to as an hypothesis. A group of hypotheses and data is called a model, and can be used to generate predictions via inductions. For example, the hypothesis that falling objects accelerate at a rate of 9.8ms^-2, combined with the datum that a given object was dropped from a height of 40m, can be used to generate the prediction that the object will hit the ground approximately 2.9 seconds later. If the prediction is disproven (for example if the object suddenly accelerates off into space) the model will be disproven, and hence one of its component hypotheses must be false.
There are two more conditions before an hypothesis can be considered a scientific hypothesis, and they concern its relation to other hypotheses. The conditions are:
1) The hypothesis must be an improvement - it must make predictions that are not made by the current best hypothesis (the CBH). Mathematically speaking, the hypothesis's predictions must not be a proper subset of the CBH's predictions.
2) The hypothesis must be parsimonious - roughly speaking, it must contain fewer "magic numbers" than the CBH. This condition only applies if the two hypotheses make the same predictions, or each makes predictions that the other does not make.
Broadly speaking, the CBH is taken to be the most predictive, parsimonious known hypothesis that has survived attempted falsification - see the Credibility section below for more detail.
These conditions are obvious if we consider a few examples:
- The "improvement" condition represents our preference for the hypothesis that Einsteinian gravity applies to all objects in our universe over the less predictive hypothesis that it applies to all objects in our solar system.
- The "parsimony" condition represents our preference for the hypothesis of Einsteinian gravity over the less parsimonious hypothesis of epicycles.
- The pre-eminence of the "improvement" condition over the "parsimony" condition represents our preference for the hypothesis "falling objects accelerate at a rate of about 9.81ms^-2" over the less predictive but more parsimonious hypothesis "falling objects accelerate".
If two hypotheses make different predictions, and neither is obviously more parsimonious than the other, then the hypotheses can legitimately be said to be competing. An example of competing predictions might be the various forms of String Theory - they make different predictions and there is no easy way of telling which is the most parsimonious. In general, the presence of competing hypotheses can be taken to mean that scientists have not been doing enough experiments. In String Theory, for example, scientists are struggling to gain funding for the massive particle accelerators necessary to figure out which is right.
If two hypotheses make exactly the same predictions, and neither is obviously more parsimonious than the other, then the distinction between them can legitimately be said to be a philosophical problem rather than a scientific one.
Credibility
The fundamental way in which a scientific hypothesis can gain credibility is by failing to be falsified. This is considered to be a strong confirmation of the hypothesis. The more experiments that are done to test an hypothesis in different ways, the more credibility that hypothesis gets. Once an hypothesis is considered by the scientific community to have passed beyond reasonable doubt, it is referred to as a theory.
There is in fact another way that an hypothesis can gain credibility. If it makes verifiable conjectures which are later confirmed, that decreases the number of alternate hypothesis that fit the data. This "weak confirmation" is roughly analogous to winning a horse race by shooting the other horses - it doesn't offer any opportunity to improve the predictive power of science as a whole - and hence is considered somewhat less desirable.
One example of this weak form of confirmation lies with the hypothesized evolution of whales from mesonychids. From this hypothesis, the verifiable conjecture could be derived that the fossil record would show intermediate forms (this conjecture is not falsifiable as you could never exhaustively check the entire fossil content of the Earth). When this conjecture was confirmed, it was considered to give added credibility to the mesonychid hypothesis of whale evolution, as it raised the bar that alternative hypotheses would have to vault by increasing the amount of data they'd need to explain.
Comparing ID and the Theory of Evolution in Biology
One issue that has been raised at some length is whether the theory of evolution itself is actually an hypothesis. The argument goes as follows: The theory of evolution states that "the diversity of life is due to evolution." This isn't falsifiable because anti-ID biologists can come up with some kind of evolutionary "just so story" to explain away any data, no matter how odd. In other words, to falsify Darwinian notions of evolution, one must demonstrate that no possible gradual pathway may have resulted in a given biological structure. Additionally, anti-ID biologists are likely to invoke ignorance to give non-telic evolution a pass on falsifiability. This is done by some polemic phrase like, “just because we don’t know how it works does not mean we should give up on science.”
This argument is in some sense correct - in general, rarefied inferences, be they of evolution or of design, are not falsifiable. However, the theory of evolution can be rephrased in the following way: "the best hypothesis for the development of any organism will only involve the evolutionary model". This is a valid hypothesis (or meta-hypothesis, if you prefer) as it can be falsified by finding one single non-evolutionary hypothesis that makes better predictions about the emergence of a biological feature or species than does the current best non-telic evolutionary hypothesis. This meta-hypothesis is considered a theory due to the apparently large amount of unsuccessful effort that people have put into falsifying it.
The ID conjecture is that “there are phenomena in the universe that are best explained as the result of intelligence.” This is eminently falsifiable, since any alternative hypothesis based on law or chance that explains a phenomena better will easily do away with intelligent design. As William Dembski explains:
- ‘‘If it could be shown that biological systems like the bacterial flagellum that are wonderfully complex, elegant, and integrated could have been formed by a gradual Darwinian process (which by definition is non-telic), then intelligent design would be falsified on the general grounds that one doesn't invoke intelligent causes when purely natural causes will do. In that case Occam's razor finishes off intelligent design quite nicely.’’[1]
The ID conjecture is also verifiable, with many theorists currently working on the task of design detection. Moreover, ID makes unique predictions. So ID will additionally be able to strengthen its place in science by finding a concrete hypothesis fulfilling these criteria. Therefore, ID is shown to be a valid scientific hypothesis.
Footnotes
- ^ From Is Intelligent Design Testable? by William Dembski
