In scientific theory, underdetermination refers to situations where the evidence available is insufficient to identify which belief we should hold about that evidence. For example, if all that was known was that 10 dollars was spent on apples and oranges, and that apples cost 1 dollar and oranges 2, then we would know that 9 apples were not purchased, but we could not know which combination of apples and oranges were purchased. In this example we would say that belief in what combination was purchased is underdetermined by the available evidence.
Contents |
Ancient Greek skeptics argued for equipollence, the view that reasons for and against claims are equally balanced. This captures at least one sense of saying that the claims themselves are underdetermined.
Underdetermination, again under different labels, arises in the modern period in the work of René Descartes. Among other skeptical arguments, Descartes presents two arguments demonstrating underdetermination.
Descartes's dream argument points out that, while dreaming, perceived experiences (for example, falling) do not necessarily contain sufficient information to deduce the true situation (being in bed). Since one cannot always distinguish dreams from reality, one cannot rule out the theory that one is presently dreaming rather than having veridical experiences; thus the theory that one is having a veridical experience is underdetermined.
Descartes's demon argument is a variant of the dream argument that posits that all of one's experiences and thoughts might be manipulated by a very powerful being (an "evil demon") that always deceives. Once again, so long as the perceived reality appears internally consistent to the limits of one's limited ability to tell, the situation is indistinguishable from reality, one cannot logically determine between correct belief from being misled; this is another version of underdetermination.
To show that a theory is underdetermined, one must show that there is a rival theory, equally well supported by the standards of evidence. A trivial example of underdetermination is the addition of an observer. For example, there is the theory that "objects near earth fall toward it when dropped". A rival theory is that "objects near earth fall when dropped but only if we check to see that they do". This rival is generated by taking any accepted theory and appending to it "whenever we look for evidence." Since one may append this to any theory, all theories are at least trivially underdetermined. If one considers such modifications of theories to be illegitimate then such "tricks" are not to be considered demonstrations of underdetermination.
More serious cases of underdetermination are illustrated when a theory admits several possibilities for which the evidence says nothing. According to Isaac Newton's mechanics, there is an absolute space in which events are located but all that can be detected are differences between velocities. Hence, it is equally consistent with this theory to say that the solar system is at rest, as it is to say that it moves at a velocity of 37 m/s in the direction from the center of the earth to the north pole. Newton himself indicated these two possibilities are indistinguishable.
Arguments involving underdetermination attempt to show that there is no reason for belief regarding some theory because it is underdetermined by the evidence. Since the evidence does not show that the theory is the uniquely true hypothesis, there is no reason to believe it rather than some equally supported rival.
Because arguments involving underdetermination involve both a claim about what the evidence is and that such evidence underdetermines a theory, it is often useful to separate these two claims within the underdetermination argument as follows:
The first premise makes the claim that a theory is underdetermined. The second says that rational decision (i.e. using available evidence) depends upon evidence that underdetermines the theory.
Any phenomenon can be explained by a multiplicity of hypothesis. How, then, can data ever be sufficient to prove a theory? This is the epistemological problem of the indeterminacy of data to theory.
The poverty of the stimulus argument and W.V.O. Quine (1960) 'Gavagai' example are perhaps the most commented variants of the epistemological problem of the indeterminacy of data to theory.
Some of the most powerful skeptical arguments appeal to the fact that all the evidence we could ever gather would still fail to determine which theory was true. It would remain compatible with 'skeptical hypotheses' like the maintenance of a complex illusion by Descartes' evil demon or (in a modern updating) the machines who run the Matrix. The skeptic argues that this undermines any claims to knowledge, or even (by internalist definitions), justification.
Philosophers have found this argument very powerful. Hume felt it was unanswerable, but observed that it was in practice impossible to accept its conclusions. Influenced by this, Kant held that while the nature of the 'noumenal' world was indeed unknowable, we could aspire to knowledge of the 'phenomenal' world. A similar response has been advocated by modern anti-realists.
Underdetermined ideas are not implied to be incorrect (taking into account present evidences), rather we cannot know if they are correct. If we presuppose that conflict between different views means that none of them are correct, this is the Middle ground fallacy.
In the philosophy of science, underdetermination is often presented as a problem for scientific realism, which holds that we have reason to believe in unobservable entities (such as electrons) talked about by scientific theories. One such argument proceeds as follows:
Particular responses to this argument attack both the first and the second premise (1 and 2). It is argued against the first premise that the underdetermination must be strong and/or inductive. It is argued against the second premise that there is evidence for a theory's truth besides observations; for example, it is argued that simplicity, explanatory power or some other feature of a theory is evidence for it over its rivals.
A more general response from the scientific realist is to argue that underdetermination is no special problem for science, because, as indicated earlier in this article, all knowledge that is indirectly supported by evidence suffers from it - for example, conjectures concerning unobserved observables. It is therefore, ironically, too powerful an argument to have any significance in the philosophy of science, since it does not cast doubt uniquely on conjectured unobservables.