Apr 13 2009

Inductive Reasoning In Science

Recently I received the following question from an SGU listener named Marty:

I’ve been debating with a friend about the nature of science, and he brought up the following argument:

“1. All inferences from experience to conclusions about the future presuppose the principle that the future will resemble the past. (Principle of the Uniformity of Nature)
a. If we suspect that the course of nature may change and that the past is no guide to the future, then all experience becomes useless and does not support any conclusion about the future.
2. Therefore, no argument from experience can support the principle that the future will resemble the past.
3. No deductive argument can establish the principle that the future will resemble the past.
4. Therefore, the Principle of the Uniformity of Nature cannot be rationally justified.
5. If the Principle of the Uniformity of Nature cannot be rationally justified, then inductive reasoning in science cannot be rationally justified.
6. Therefore, inductive reasoning in science cannot be rationally justified. ”

Your thoughts?

This type of question comes up frequently – they essentially are attempts to use philosophy to argue that science cannot lead to objective truth, therefore science is not valid (or at least I can ignore it whenever I choose, which is typically how such arguments are applied). The problem with all such arguments is that science is not about objective metaphysical truth, but rather it is a collection of methods for making abstract models of nature and then testing those models against reality.

Inductive Logic

First a word on inductive reasoning, which is one of the types of reasoning in science (but not the only one). Induction is the process of going from the specific to the general, or forming a conclusion about the nature of the universe from a limited set of observations. One classic example is the fact that, so far, it has always been observed that the sun rises each day in the East. Therefore we can infer that the sun always rises each day in the East.

Induction is distinguished from deduction, which can be summarized as going from the general to the specific. If we take as a premise that the sun always rises each morning in the East, then we can deduce that the sun will rise tomorrow morning in the East. In valid deduction, if the premises are true then the conclusion must be true.

Induction is different – the observations may be true but because they are limited the conclusion may still be false. The classic example here is the observation that all swans ever observed are white, leading to the conclusion through induction that all swans are white. This was a reasonable conclusion until black swans were discovered in Australia.

Induction and Science

Science certainly involves induction, but it is not limited to it. But because much of scientific reasoning is inductive that has led to the philosophical question of how valid are conclusions in science. This is not a new question, and Marty’s friend would do well to investigate some of the extensive discourse on this question.

Philosopher Karl Popper had an interesting answer to this question – inductive reasoning does not exist, and therefore science is not induction. Rather he focused on verification and falsification. He argued that science comes up with hypotheses and theories that make predictions and therefore can be tested. He also noted that there is an asymmetry to this in that thousands of verifications cannot prove a theory correct, but one falsification can prove it wrong. Therefore the ability to be falsified is a necessary feature of any truly scientific idea.

This also means that no theory can be absolutely validated – only tentatively validated. Therefore science never arrives at absolute certainty. There may always be a black swan to be discovered out there. But scientific induction can lead to conclusions that have been validated to such a degree that we can comfortably act as if they are true. I don’t think anyone should waste any time or resources preparing for the possibility that the sun will not rise tomorrow.

Much time has also passed since Popper, and his ideas have been greatly extended. I will ignore for this post what I consider to be a huge diversion into post modernism but rather fast forward to a more contemporary vision of the nature of science.

While Popper, in my opinion, was essentially correct his view was incomplete. For example, science does not only consider one hypothesis at a time. Rather collections of theories are evaluated together over time by a community of scientists. Each theory must not only survive falsification it must also be consistent with other established theories – it must fit into the web of evolving scientific theories.

There are also other processes to consider in science, such as causal inference. Science does not just describe what exists or what happens in nature, it tries to explain how things happen – what causes what. For example, we may investigate genetic causes of the color of swans, or even the effects of various pigments on the absorption and reflection of various frequencies of light and how they affect the cones in our retinas.

Causal inferences often also require triangulation from multiple independent lines of evidence. Evolution, for example, cannot be established by one line of evidence, but requires many – the fossil record, genetic homology, developmental biology, population genetics, etc.

The Principle of Uniformity

The reasoning employed by Marty’s friend does not seem to take into account the actual nature of scientific conclusions. What they are saying, essentially, is that science does not yield philosophically certain truth – conclusions that must be true. This is correct but irrelevant, since that has never been part of science. Rather science comes up with models of nature and then tests those models. It retains those that are validated, but only for as long as they are validated, while continuing to try to falsify even previously validated ideas. It rejects or modifies falsified ideas. Scientists also try to infer causality in order to develop an increasingly complex abstraction of nature which makes predictions that can be further tested. And all the independent threads of scientific ideas must weave together into one seamless web.

The notion that there is temporal uniformity in nature – that the future will resemble the past, is not really a premise of scientific induction but rather one thread in the tapestry of science. It itself is a hypothesis that makes predictions and can be tested. For example, we have been measuring the speed of light very accurately for decades. If the speed of light were not constant and changed over time we could detect it. Therefore the constancy of the speed of light is falsifiable, but so far has been verified.

The same is true of all constants and laws so far discovered in nature. The laws of nature do not appear to change over time, and they also appear to exist throughout the universe. The principle of uniformity has been verified so far.

The only refuge for someone denying the utility of scientific induction and scientific reasoning more generally is to say that nature is so capricious and inscrutable that we cannot even reason about the principle of uniformity, or any other basic law or constant. However, such arguments, as the one above, generally take the form of using logic to demonstrate that science cannot reach conclusions that must be true – logically, 100%, metaphysically true. They then conclude that science is not valid, or “cannot be rationally justified.” But this is a false premise, because science has never been about logical truth, as I described above.

The real question is – is science pragmatically valid. Does it do what it claims to do. Here we have the metaexperiment of science itself. Over the last few centuries of formalized scientific investigation, what has science produced? If nature were inscrutable and the laws and constants that we infer from it of no utility, then science should not have progressed much or at all over the last few hundred years.

And by progress I do not mean just the ability to weave explanations of how the world works – any system can potentially do that. Rather I mean – has science given us an increasing ability to make predictions about nature that are later validated. Here the answer is clearly yes. If the principle of uniformity and all of our abstractions about gravity, mechanics, electricity, and the structure of the solar system were of no utility, then we could not send a hunk of metal from the earth and guide it to a distant planet and receive as reward stunning pictures of Saturn.


I may not be able to prove philosophically that the sun must rise tomorrow, but I can infer from observation and induction that it is overwhelmingly probable that the sun will rise tomorrow (and I can even predict when and where). As Stephen J. Gould wrote:

In science, “fact” can only mean “confirmed to such a degree that it would be perverse to withhold provisional assent.” I suppose that apples might start to rise tomorrow, but the possibility does not merit equal time in physics classroom.

34 responses so far