The Limits of Science
Lessons in the Philosophy of Science
In the evidence-based fitness space, we often worship science as an enterprise.
The more intellectually prone in the space will often talk about science in terms of method and that the method of science is what distinguishes it from other endeavors.
Consider the scientific process: a theory is proposed, and the proposers make predictions about their hypothesis, which are then tested. The findings of that research are then subject to review by peers in the field. If the results have been represented clearly and transparently, the methods used in the study to derive a particular result will be critiqued to determine whether they do in fact support that outcome.
But the method of science or even the “process” of science is limited in its philosophical implications. I will outline three main issues that limit science and our understanding of it.
These issues are:
The Problem of Demarcation
The Problem of Induction
The Is/Ought Problem
So, let’s cover each problem and how they potentially limit science.
The Problem of Demarcation
One simple but fundamental problem is… what separates science from non-science? Or put in another way what would separate science from journalism, astrology, painting, or creative writing?
What makes science distinct? How is science different from pseudoscience?
Now, this might sound simple and easy to answer, but when you think about it this becomes a complicated problem.
First, what do we define as science? If we were to define science by a set methodology or process we quickly run into issues.
The methods of one scientific discipline, such as physics, might not apply to another discipline like psychology. In addition, the ‘social’ sciences are unique not only in method but in aim and practice.
For example, branches of physics can rely on complex mathematical computations to make predictions about the natural world. While psychology often relies on a set of observations between complex human beings to find differences and then justifies those differences via statistical methods to describe a universal ideal of human behavior.
It is challenging to establish science via method for reasons beyond what I’ve highlighted, let’s pretend we can separate science from pseudoscience by a set of criteria.
What criteria would we use?
If we use philosopher Karl Popper’s criterion of falsification this would mean the ability to test a hypothesis and subsequently disprove it via that test would separate science from pseudo-science.
For instance, we would consider Einstein’s theory of relativity scientific because his theory was tested in light of observations made during a solar eclipse, it had the potential to be shown false. This is unlike Freudian psychoanalysis which could never be proven false as it could explain away any of its shortcomings.
Explanatory power does not make a theory scientific according to Popper, merely the ability to be put to the test and shown to be possibly false. The theory which survived various tests and was shown not to be false could be seen as more likely the case.
But even falsification has its problems… namely two major ones.
Requiring a strict criterion of falsification might rule cases of clear science as non-scientific. e.g. the theory of evolution
Even pseudoscientific practices could be considered science if they make testable claims, regardless of whether they are false. e.g. astrology
So, if we cannot rely on falsificationism, what other criteria can we rely on?
Rather than coming up with another proposed answer by philosophers let’s cut to the chase… Any criteria philosophers propose for the demarcation problem are insufficient for various reasons. The best we can do is come up with characteristics that define science versus other endeavors, sort of like family resemblances.
Note that this is a major limitation for science, we have no clear demarcation of science yet.
The Problem of Induction
Outside of deductive inferences, how do we typically reason?
Think of deduction in terms of propositional logic, each premise leading to a conclusion is necessarily and sufficiently true. Propositional logic relies on deductive reasoning which is always certain.
We rely upon inductive inferences because the premises and conclusions we reach are never certain.
This is where we get concepts such as abductive reasoning from and on the surface it makes sense to suggest we reason this way in science. We would hope that scientists use some form of inference to the best explanation in their thinking, but are scientists justified in relying upon these inferences?
Philosopher David Hume argues that we cannot rely on induction to make justifiable claims to knowledge.
We can formulate Hume’s argument this way (this formulation comes from philosopher Michael Huemer I believe):
1. Our beliefs can be divided into three categories:
(a) Beliefs about relations of ideas: These are beliefs that are true by definition
and can be known independent of observation, including (Hume thinks)
such examples as “All bachelors are unmarried” and “2 + 2 = 4.” In modern
terminology, these would be called “analytic a priori” beliefs. (b) Beliefs about observed matters of fact: These would be things one is actually
observing or has observed. (Matters of fact include all things that are not
relations of ideas.) For example, my belief that there is a desk here (while I
am looking at it and touching it) is an observed matter of fact.
(c) Beliefs about unobserved matters of fact: These include all matters of fact
that I am not observing and have not observed. For instance, my belief that
there are people living in China is an unobserved (for me) matter of fact, as
is my belief that the sun will rise tomorrow. (Tomorrow, assuming I witness
the sunrise, it will become an observed matter of fact.)
2. All unobserved matter-of-fact beliefs depend upon inductive inference for their
justification. For instance, I think the sun will rise tomorrow because it has risen
in the past, and I am generalizing on that experience. I think there are people in
China because I have heard things about the people in China and I believe the
testimony of others; but I believe the testimony of others because I have formed
an (inductive) generalization that people usually tell the truth (in this sort of
circumstance, about this sort of thing).
3. All inductive inferences presuppose some such premise as “The course of
nature is uniform” or “Unobserved things will resemble observed things.” Call
this “the Uniformity Principle.”
4. The Uniformity Principle is not a relation of ideas proposition, since it is not
analytically true. It is logically consistent to hypothesize that the course of nature
may not be uniform.
5. The Uniformity Principle is not an observed matter of fact, since it makes a claim
about unobserved objects.
6. The Uniformity Principle is an unobserved matter of fact belief. (From 1, 4, 5.)
7. The Uniformity Principle depends for its justification on induction. (From 2, 6.)
8. But the Uniformity Principle cannot be justified by induction, since all inductive
inferences presuppose the Uniformity Principle (premise 3), and circular
reasoning is not acceptable.
9. The Uniformity Principle cannot be justified. (From 7, 8.)
10. No inductive inference can be justified. (From 3, 9.)
To simplify this, induction depends on a fundamental assumption according to Hume, namely, the uniformity principle. The uniformity principle suggests that nature stays constant or “uniformed” and thus we can use inferences to make valid predictions in the world.
But the problem comes in when we try to justify the uniformity principle itself. We would need to appeal to induction in order to justify the uniformity principle which would then be used to justify induction… the reasoning becomes circular and this is bad.
This means we have to reject the uniformity principle which means induction is unjustified. In practical terms, science cannot depend on inductive inferences to justify any claim nor can it make predictions. We cannot infer future events or outcomes from past observations if we cannot show that nature remains constant.
Some philosophers, like Popper, tried to overcome this problem by rejecting induction entirely for falsification. But as we saw falsification has its own issues.
Currently, no philosopher has overcome this problem.
The Is/Ought Problem
Another problem brought to us by David Hume is the is/ought problem.
Hume argues that we cannot derive moral statements (prescriptive statements) from fact-based statements (descriptive statements).
This means we cannot infer what we ought to do in any given situation by what is. This means facts, even those gathered from the sciences, cannot tell us what to do when it comes to ethics.
Morality or ethics is distinct from facts about the world.
For example, knowing what trimester a fetus develops a heartbeat in does not tell you whether abortion is morally justifiable.
Some moral realists would argue that there are moral facts making this distinction pointless, but the establishment of what constitutes a moral fact is contentious. So far, we do not have a solid way to overcome this problem.
Science can describe the world well, but it cannot tell us or inform us about what to do with that information.
Conclusion
I am a person who is very much pro-science, and so are many philosophers, which is why they dedicate themselves to these questions to find out what makes science so special.
However, we need to acknowledge these pressing problems in our fundamental understanding of what science is and what it does.
So far, we do not have a clear demarcation of science, we probably can’t rely on inductive inferences for making predictions both observed and unobserved about the world, and we probably cannot use science to deliberate in other ethical dilemmas.
Truly understanding and appreciating science would mean taking into consideration its strengths and limitations.