Scientific Method: Some Clarifications
The Problem of Induction
One of the aims of science is the search for general statements, or ‘laws’, that explain the world around us. The dominant tradition of how to engage in science has been empiricism. This philosophy states that knowledge is exogenous — by accumulating facts about the world, we can find regularities and laws. So there’s strong emphasis on experimentation/observation, meticulous recording and accumulation of data, and induction. This is the method of basing general statements on accumulated observations of specific instances. Induction was regarded as the dividing line between science and non-science.
But we can’t logically derive general statements from specific observations, no matter how many instances of singular observations we accrue. Consider the rising of the sun. I’d bet my life that the sun will rise tomorrow, but I can’t logically prove it by pointing to a huge number of previous instances in which it did so. Here we see a gap between psychology and logic. Why is that? You can point out that, ‘But we can predict with great precision the exact time of the rising of the sun for tomorrow.’ You’d be right. But you’d be begging the question. You’d be using general statements to prove induction. Plus, the problem of logical entailment would still be standing. Because of this, most people just pretend that we don’t have a problem here and take induction as a given. But then they can be easily criticized by people who point out that the foundation of this whole enterprise is a sham. This leads some people to irrationalism and mysticism.
Isn’t the goal of science to generate extremely precise predictions of certain future events? Is science a collection of laws that faithfully predict future events, or is it our best current explanation of the natural world? The latter sounds a bit vague, so most people stick with the former definition, while sweeping the problem of induction under the rug.
Naive Falsificationism or Misunderstanding Popper
Logically, we can’t derive a general statement from many confirming instances of an event/experiment, but we can definitely refute a general statement by a single observation. This is the asymmetry between verification and falsification. But bear in mind that this is only a matter of logic. In practice, in the domain of methodology, we can always question the observation/experiment that is supposed to refute the general statement. We can inquire about the collection of data, the manipulation and processing of data, and even the motives of the persons involved. Most people think that this is a refutation of Popper’s falsificationist approach. But he didn’t propose ‘falsificationism’ as a logical framework that should be followed step-by-step. Popper was against all such methodologies.¹ His whole epistemology rests on the assumption that all knowledge is theory-laden: observation by itself is meaningless without an underlying explanation. So Popper wasn’t as naive as every philosophy course claims he is. Popper’s formulation of falsificationism emphasizes that experiments are not executed in order to refute an entire theory, but rather to find a problem within a theory.² This is the critical approach, a way to find faults with our theories. Only in this way can we make progress.
Verificationism is essentially a process of confirmation bias. You can always find some evidence for any hypothesis. Instead of looking for massive amounts of confirming events for a theory and building a good story, we should give precedence for verifications that stem from the theory deductively. Verification should count when you have a theory, derive a prediction from that theory, and then test it. These kinds of experiments are crucial, because they can show that there’s something wrong with your theory. This is what Popper meant by falsificationism.³ Experiments can be useful either for testing predictions that follow logically from theory or for deciding between two theories.
The Source Does Not Matter
Sometimes people talk past each other on the subject of induction because they confuse psychology with epistemology. People can be quick to point out that some person had a singular observation from which they have constructed a hypothesis. Whether we come by hypotheses in our sleep, or derive them from reading literature or ancient myths, or from some observation, is irrelevant. There are no logical rules that mandate you how to create hypotheses and ideas. We can only criticize theories according to their content. Epistemologically, origins of theories don’t matter, though they may be of interest for psychologists and sociologists.⁴
All Observations are Theory-Laden
The idea of unbiased-by-theory pure observation is a myth. Observation is necessarily discriminatory. You can’t observe everything. Meticulous collection of data is specifically constrained by some ideas. A biologist may witness something in the course of an experiment that she would regard as discovery, because she already has background knowledge against which she can interpret the chance observation.
A Case Study by Bryan Magee
Consider the hypothesis that water boils at 100°C. Logically, we can’t prove this hypothesis by testing it a million times. What about credence — the more confirmation tests of this hypothesis we have, the more we’re willing to believe it. But this is the wrong approach. We should instead proactively try to find problems with the hypothesis. When we take the problem-seeking approach, we draw from the well of creativity. We could try to boil the water in a closed container, and this way we would discover that our hypothesis isn’t universal. We would then narrow our hypothesis to ‘water boils at 100°C in open containers’. We could test our new, narrowed-down hypothesis at high altitude. Here we would discover that water has a lower boiling point at high altitudes, and we would further refine our hypothesis as ‘water boils at 100°C in open containers at sea-level atmospheric pressure’.
As Bryan Magee said: “For when we discovered that water did not boil at 100° Centigrade in closed vessels, we had our foot on the threshold of the most important kind of discovery of all, namely the discovery of a new problem: ‘Why not?’ We are challenged now to produce a hypothesis altogether richer than our original, simple statement, a hypothesis which explains both why water boils at 100° Centigrade in open vessels and also why it does not in closed ones; and the richer the hypothesis, is the more it will tell us about the relationship between the two situations, and the more precisely it will enable us to calculate different boiling points.”⁵
Progress depends on the presence of problems. Searching only for supporting evidence stifles progress. Instead, criticism is required. Seeking confirmation of one’s hypothesis is predicated on the false assumption that by amassing many confirmatory instances, one is justified in accepting it. But we can never achieve certainty. All we can do is conjecture ideas and criticize them. Those hypotheses that withstand criticism are regarded as our best explanations of the world until when we come up with better ones.
- Popper, Karl. The Logic of Scientific Discovery (p. 31)
- Hall, Brett. Philosophy of Science. (http://www.bretthall.org/philosophy-of-science.html)
- Popper, Karl. Conjectures and Refutations, Science: Conjectures and Refutations (pp. 47–8)
- Popper, Karl. The Logic of Scientific Discovery, The Problem of Induction
- Magee, Bryan. Popper (Fontana Modern Masters, 1973)
Mizrob’s interests include evolutionary biology, epistemology, ethics, history, and everything else. Follow him on Twitter @mizroba.