Determinism and free will (1/2) (translation in progress 🙂 )
Scientific realism and constructive empiricism (2/2) (translation in progress 🙂 )
Scientific realism and constructive empiricism (1/2) (translated)
Welcome to this blog ! (translated)
Scientific realism and constructive empiricism (1/2)
In this first post, I am going to speak about what is called « constructive empiricism ». The appellation can look scary but it is perfectly intelligible by going step-by-step !
Let us start from the beginning. If we take a step back from scientific theories, we can wonder to what extent we can trust what they say. Do they describe, at least approximately, what really happens in reality ? Do unobservable entities like atoms or electrons really exist ?
All these issues relate to an ubiquitous debate in philosophy of science: this is the problem of « scientific realism ».
Scientific realism is an optimistic view on the capacity of science to tell us things about the world. For more clarity, this stance is very often divided in three parts :
1) A « metaphysical » aspect : there exists a mind-independent world. If the human species were to disappear, trees, houses, mountains, galaxies, in others words everything that surrounds us, would still exist. It can seem obvious but it is worth mentioning it !
2) A « semantic » aspect which concerns the way to construe scientific theories : there are to be « construed literally ». It means that scientific terms aim to describe the natural world in all its aspects : not merely what we can observe in everyday life, but also what is unobservable for humans. For instance, terms like « electron » or « atom » seek to refer to microscopic entities of the world.
3) An « epistemological » aspect concerning the nature of scientific knowledge : our best scientific theories describe, at least approximately, what actually happens in reality. They are « approximately true ». In particular, we have good reasons to think that unobservables like electrons or protons really exist.
Scientific realism appears to be very natural at first sight. When we are taught chemistry, biology or physics in school, we do not care about the question whether DNA, atoms or protons exist or not : it seems to be an obvious presupposition.
But it has to be said that there are a lot of arguments that put into question this vision of science. Instead of listing all these arguments, I am going to illustrate some of them by presenting a view which stands in contrast to scientific realism : « constructive empiricism ».
What is constructive empiricism all about ?
This position was introduced by Bas Van Fraassen in his book The Scientific Image, in 1980. Whereas the philosophical discussions tackled in majority the « semantic » aspect of scientific realism during the first part of the twentieth century, Van Fraassen refocused the debate on the « epistemological » aspect. His now-famous theory has been subject of many debates that I will try to explain here.
So, to what extent is Van Fraassen opposed to scientific realism ? Why is he an « anti-realist » ?
Although he completely agrees with the first two claims of scientific realism, he refuses the third aspect. According to him, « Science aims to give us theories which are empirically adequate ; and acceptance of a theory involves as belief only that it is empirically adequate » (p12). A theory is empirically adequate if it « saves the phenomena, that is that it correctly describes what is observable » (p4).
Thus, constructive empiricism is less optimistic than scientific realism regarding the epistemological status of scientific theories. Accepting a theory merely entails to believe in the truth of all that it can possibly say about the observable world. We should remain agnostic about all propositions that concern unobservable entities, like for example « Electrons exist ».
Before clarifying Van Fraassen’s motivations for introducing such a position, it looks necessary to precise what he exactly means by observable and unobservable. According to him, an object is called observable if a human can observe it with the naked eye, at a certain time and location. Observable phenomena can therefore be past, present or future. It is important to distinguish between « observable » and « what has been observed so far ». For instance, the next solar eclipse is clearly an observable phenomenon, but it has still not been observed.
However, one may wonder if this distinction is not simply ill-defined. All humans have not the same vision capabilities : the differences between a blind person, a newborn and an adult with perfect eyesight are evident. The notion of « observability » appears to be completely subjective. Van Fraassen acknowledges that there is no sharp line between observables and unobservables, but he argues that it can remain a relevant differentiation. As every measuring instrument, human eyesight has limits, so there are cases where the distinction is unambiguous : for example, everybody agrees to say that a tree is « observable », whereas an electron is « unobservable ».
We are now ready to dive into the heart of the matter !
So, what are Van Fraassen’s arguments against scientific realism and why does he consider his theory to be better ?
I am going to present the first criticism of Van Fraassen : it involves the problem of « undetermination of theory by data ».
Undetermination and relation between theory and experience
Let us imagine an experiment, where you measure with a speed camera the speed of a car on a highway. You obtain the result : 2km/h. What is your reaction ? You probably say « My speed camera does not work ! », and you are right !
This little example illustrates that when one has a theory like « The speed of a car on a highway is high », one cannot deduce the « prediction » that the speed camera will show a high speed. To do so, one has to suppose that the speed camera is not defective, or that the environmental conditions do not disturb its functioning.
Let us generalize a bit : if one has a theory « T », one must admit that T is not sufficient to make any observable predictions « O ». In order to do so, one has to appeal to additional hypotheses « H1, H2, H3… » on the operation of measuring devices, experimental conditions, etc : these are called « auxiliary hypotheses ».
Now, if one obtains a contradiction between observable predictions and experimental results, what can be concluded ? As illustrated in the example with the speed camera, one cannot deduce that T is false. Actually, it is the whole set of assumptions « T and H1, H2, H3… » which is false. So either T is false, or one of the auxiliary hypotheses is false, but one cannot exactly know where the error is. This is called « confirmation holism » : T can never be tested in isolation from background assumptions. Scientific theories are always tested as a whole.
Accordingly, to make the set « T and H1, H2, H3… » consistent with the observations, there are several possibilities : change the theory T, or change/add/remove auxiliary hypotheses.
This entails what is called undetermination of theory by data : giving any set of experimental data, it is always possible to have mutual incompatible theories (in particular in relation to unobservable aspects) that are all consistent with the data.
But in this case, how to choose between these « empirically equivalent » theories in order to make scientific progress ?
This is where various non-empirical criteria come into play, like simplicity, coherence with the rest of our knowledge, elegance, explanatory power, or fruitfulness. Scientific realism proponents appeal to these extra criteria by making an « inference to the best explanation » : theories with great simplicity or explanatory power are to be preferred, because they are more likely to be true than other empirically equivalent theories. This method would enable us to progress towards truth about observables as well as about unobservables.
Argument against scientific realism
But the problem is, it is hard to see how such non-empirical criteria could be truth-indicators. As Van Fraassen states : “it is surely absurd to think that the world is more likely to be simple than complicated” (p90).
That is why it seems more reasonable and careful to be agnostic about unobservable aspects of scientific theories. In this context, we can explain theory choices by saying that non-empirical factors are merely pragmatic virtues : the aim of science is to correctly describe all observable phenomena in the most coherent, simple and intelligible way.
Of course, when we commit to the truth about observable aspects of a theory, we go beyond the evidence and do not solve completely the underdetermination problem : there are several theories that can be compatible with what has been observed so far but that will disagree regarding future observable but not-yet-observed phenomena. However, by limiting our epistemic commitments to observables, constructive empiricism allows us to take much less risks compared to scientific realism. It therefore seems to be a way better alternative !
That is all for this first blog post ! We will continue in the next one with other arguments against scientific realism. We will also see that at the end of the day, constructive empiricism is not that compelling !
Welcome to this blog !
Welcome to « Science Peak » (or should I say « Science Speak » ?), where I will speak about topics around philosophy of science.
How to define the scientific method ? What distinguishes science from pseudoscience ? What do scientific theories tell us about the world ? These are all fascinating issues that I will try to address here.
(Do not hesitate to let me know if there are any English mistakes, my English is currently far from being perfect !)