Induction: a problem or a justifiable principle?

Hume recognises a dilemma in the process of induction: that it cannot be justified. Induction is a process of reasoning where ‘instances of which we have had no experience resemble those of which we have had experience’. Hume’s problem is that the process of induction can neither be justified deductively (to work from the general to the more specific: a top-down approach) nor inductively (to work from the specific to the general: a bottom-up approach).

The deductive argument would be confined to a closed domain of discourse, as the premises of the argument would be about past experience only. The conclusion that is based on these premises would not grow beyond the past and would thus preclude any possible experience of the future. The inductive argument would presume that what was good in the past will be good in the future. This, however, begs the question by assuming what is to be proved.

There are various arguments that aim to justify the problem of induction. The first of these attempts to avoid the problem of induction by showing that induction is separate to scientific method. The second of these seeks to justify the problem of induction with either a pragmatic approach or an inductive approach. The third of argument aims to eliminate the problem of induction all together by stating that the process of induction speaks for itself and need not be justified. Each will be dealt with in turn.

Avoiding the problem of induction in science

Karl Popper proposes that induction is completely independent from, and separate to, scientific method. Popper reasons that, while on the one hand, inductive arguments always lead to invalidity; on the other hand, empirical scientific methods always lead to validity. Popper makes a distinction between inductive arguments and empirical scientific methods by an explanation of falsification. He states that science can be described through hypotheses that explain the work of the world, rather than any form of absolute truth. He further argues that the more a hypothesis survives severe tests of falsification; the better a hypothesis is corroborated.

Through tests of falsification, if the hypothesis deduces a false observation, that hypothesis will be disconfirmed. By contrast, if a true observation is enunciated from the hypothesis, then it will be confirmed. In this respect, failing to falsify the hypothesis equates to proving the hypothesis is positive. For example, if you make the statement that ‘all magpies are black’ – you are, in effect, stating that you have never observed a non-black magpie. Popper would suggest that, due to this statement surviving tests that aim to falsify it (such as searching for magpies that are non-black), the statement is corroborated. While this theory is logical in and of itself, by accepting that statements can be corroborated through the process of falsification, this raises the issue of induction: an instance of which you have had no experience (nonobservance of a non-black magpie) resemble instances of which you have had experience (observing only black magpies). Accordingly, Popper’s approach fails to prove that scientific method operates independently from, and separately to, induction. Rather, it shows that they are inextricably connected.

Justifying the problem of induction: the inductive argument versus the pragmatic justification

The inductive argument proceeds on the footing that if something worked well in the past, that same thing will work well in the future. Max Black notably points out that this approach does not presume that the conclusion about the future is a premise. Rather, the conclusion about the future (in our case, that something will work well in the future) is an inference. Nevertheless, this reasoning is circular because, even though the conclusion is not perceived to be a premise of the argument – the conclusion asserts the validity of the antecedent (in our case, that something did in fact work well in the past). This point becomes clear when considered in the context of a deductive argument: such arguments can establish the truth of its conclusion only if the premises are true and the argument is valid. The validity of the past event is asserted by the conclusion of the argument. This is fallacious because inductive arguments cannot operate to confirm their own legitimacy. Accordingly, an inductive argument will never reach a true conclusion.

Hans Reichenbach argues that, rather than using an inductive argument to justify induction, a deductive argument should be used through a pragmatic justification. By adopting deductive reasoning, Reichenbach contends that, although induction is a method of prediction – it is the best method available. Therefore, if any method can be used – it should be induction; not to prove the truth of what is asserted, but rather, to prove its reliability. To this end, Reichanbech gives two options in a hypothetical: that nature is uniform or that nature is not uniform. If it is not uniform, he argues that no method will be reliable. If it is uniform, he argues that induction is the only method that will be reliable. The essence of this argument is that we have nothing to lose and can only gain by adopting the inductive method. Notably, Reichenback’s argument does not aim to prove that nature is in fact uniform. Nor does it seek to prove that induction will always work. Rather, his argument reveals that the best method of prediction is induction – whether it works or not. His focus is to reveal best practice – without having to necessarily show that best practice is workable. In this respect, the approach does not attempt to validate but rather vindicate the inductive procedure.

Eliminating the need to justify the problem of induction

What has been described as a linguistic approach, attempts to prove that the process of induction should be accepted for what it is and there is no need to justify it as a problem. This approach rules out the need to explain why the deductive argument or the inductive argument is the better justification. It contends that induction is an independent mode of argument, just as deduction is so. It does not make sense to demand justification for induction by calling into doubt the machinery that is used for justification. Arguing that a particular type of ‘argument’, in and of itself, is not rational – is irrational. Moreover, there are different approaches to reasoning and one sense of reasoning is not necessarily better or worse than another sense. In our case, the dilemma is this: some think rationality consists in the use of counter-inductive arguments, while others understand rationality to be the use of inductive arguments. There is no way to argue that one approach is supreme over the other.

Conclusion: induction as a problem or principle?

In conclusion, there is no uniform approach to justify why past experience gives a reliable explanation for future experience: the problem of induction. On this basis, the rationalist would argue that more issues arise by recognizing induction as a problem rather than a principle. Predictions and generalizations made on the basis of induction should be accepted, as logical principles presupposed by logical proofs upon which there is no basis to reject inference. This is a more realistic approach, because justifying induction requires justification of the principle of uniformity of nature, which is something so fundamental and almost impossible to justify. Therefore, if any view must prevail against Hume’s – the rationalists view is most persuasive.