Theories with expiration date

The way of doing science has changed during history, and the process continues. If science is steadily evolving, what happens with the theories that were built using more “primitive” versions of science? Can we still use them to build our own theories, or are they dismissed? While the correct answer is probably that it is situational, there is no objective way to decide which theories are still useful and which ones are outdated, because the criteria to make this decision also evolves with time. A bit recursive, isn’t it? What is more, it opens the door to think that all the theories that we produce during our scientific life will be substituted for improved versions, if not proven wrong.

While I’m happy to embrace all these challenges of science (shouldn’t an objective construction of knowledge be objective?), I do my best to produce the best theories. In order to do so, it is important to be aware of the limitations of the conceptual framework that determines how we do science. There are a few points that I’ll briefly address: which kind of biases do we have? What is a “fact”? What does a statement require to be a scientific theory?

Being aware of the biases of your research means to understand that you are doing science at a specific place and time, with particular historical and cultural idiosyncrasies. I will name two tightly connected biases that I think affect the (neuro)science nowadays: Disproportionate mechanicism and reductionism. The neuroscience of this century has inherited the mechanistic approach of the 20th century molecular biology (A new biology for a new century from Carl Woese is an outstanding lecture on this topic1). This mechanistic approach, which can be traced back to Francis Bacon and his Novum Organum biases the scientist towards deconstructing the matter of study into smaller pieces to study them independently. Nevertheless, it is not trivial to predict how the different pieces of the system will interact among them. As Hopfield said “When we say that biology is a complex system, it is really this immense amount of information necessary to specify the significant state of biological matter (compared to an equivalent mass of geological matter) is being referred to”2. It is necessary to study the system as a whole to understand it, because the interaction between the isolated parts are too complex to be predicted.

Such mechanistic framework leads to a reductionist approach on which puzzle-solving questions about particular pieces of the systems are the mainstream. This, in turn, has led to an engineering perspective, moving from studying the pieces to manipulating them, but never focusing on the main system; or in Woese’s words “A society that permits biology to become an engineering discipline, that allows that science to slip into the role of changing the living world without trying to understand it, is a danger to itself. […] Today more than ever we are in need of a science of biology that helps us to do this, shows the way. An engineering biology might still show us how to get there; it just doesn’t know where ‘there’ is”1.

Another problem when doing science is the definition of “fact”. Feyerabend wrote “a clash between facts and theories may be a proof of progress”3; now, I wonder what is a fact. When I say that the atomic weight of hydrogen is 1 or that there is a phase alignment of the gamma oscillation of brain regions X and Y, are both facts? Even if we can consider both as facts in a broad sense, I think that we have to make a careful distinction when using them to build/challenge theories. I particularly like the differentiation between terms of first and second intention made by William of Occam. Using my own words, terms of first intention are direct descriptions of nature while terms of second intention are built on top of them. For instance, while the description and detection of an action potential can be straightforward (this is not always true but let’s idealize for the sake of the explanation), when we record some actions potentials and work with them, multiple subjective decision will be required to obtain a result, such as the window size to bin a histogram, the time-frequency tools to study the oscillatory properties of the signal or the dimensionality reduction techniques that we use, among others. It is my opinion that the facts we obtain are terms of second intention. It is obvious that we can use this kind of “facts” to do science; but it is important to understand that all the results we obtain are not completely objective measures of nature, but will be biased by previously existing scientific paradigms. What is more, the more elaborated the fact is, the more historical biases and assumptions it will suffer from. Therefore, when we are going to add/modify a theory of an existing research program4, it is important to do the mental exercise of carefully considering the premises that were used to create the facts that are going to be the base to build our theories. It’s recommendable to confirm that the floor is firm before building a house, even if we take it for granted.

At last, it’s worth to think about what makes a theory to be scientific: It must be explanatory, falsifiable and have predictive power. In other words, to produce knowledge, a theory needs to explain a topic, to be subject of experimental testing, and produce predictions of things we previously didn’t know and that can be addressed scientifically. Therefore, it is theory building and not experimental work what ultimately makes the scientists what they are. While reproducible experimentation is one of the cornerstones of science, a mere enumeration of results is not a scientific theory; in this respect Lakatos defended that “A proposition might be said to be scientific only if it aims at expressing causal connection”4. This explanatory capacity ultimately leads to the creation of new predictions (excessive knowledge) that can be experimentally tested (the theory is falsifiable). In this last point, it’s worth to note that the falsifiability of a theory changes with time; a theory might not be testable at a given time, but this can change with novel technological developments. That case is what Russell would consider a philosophical theory that in the future may move to the field of science.

I have recapped what I think are three relevant points in the construction of science. I’ve chosen these specific points to outline simply because I spend a lot of time thinking about them, but there are many others. We have to start by being aware that science is a human construct in continuous evolution, noticing the biases of science at the present time. Next, we have to think about the facts that are present in our field of research, how many assumptions do they have behind, whether we agree with them or not, and how strongly the present research programmes rely on them. At last, our work has to produce predictions that will light the path of future research, rather than providing plain mechanistic enumerations of facts. In summary, if we aim to be good scientists, we should take three premises into account when producing theories: explicability, falsifiability and predictive power.  

References:

1. A new biology for a new century.

2. Physics, Computation and Why Biology Looks so Different.

3. Against Method.

4. The methodology of scientific research programmes.