Monday 9 June 2014

Science is NOT what scientists do


I was going to keep this for the book I'm writing about Foundations of Science, but I have just read a blog post from Sabine Hossenfelder and I could not stay quiet:

Is Science the Only Way of Knowing?

I will have to respectfully, but harshly and completely disagree with almost everything she says in the article. To start with, she says the dreadful sentence that I heard many times:

"Science is what scientists do."

This is a definition that not only is not true, but also useless. Let me justify it. Let us suppose that we accept this as a definition of science. Where do we draw the line? I mean, is everything that scientists do science? When I (I am considered a scientist, by the way) fry an egg, is that science? As some can argue that it is, when I swear because something fell on my thumb, is that science? I am not exaggerating, I'm just showing that there should be a specification of what part of what scientists do is science.

Let us now assume that we can somehow say that some things are not science. Consider Sir Isaac Newton. Was it science when he was trying to draw a map of Hell? When a Nobel Prize winner starts to talk seriously about paranormal things, is that science?

You might argue that we can decide what is science by consensus, but that is more arbitrary. Consensus is a logical fallacy and although it might be evidence of the right direction, it's never a proof. Let us say that somehow most of the scientific community becomes corrupt (that can happen, just consider governments...). The leaders then start to decide what is science and what is not according to what THEY do. Those who don't agree, are not doing science by convention. If you are still doubt that this can happen, I suggest you to look for articles about the present situation in academia in the UK. Check authors like Stefano Colini and Thomas Docherty.

There is another catch: Who decides who will be a scientist? Other scientists, of course. How? By consensus. But how is this consensus achieved? Well, they need to agree on a minimum of knowledge and skills that the person needs to have. They need to agree on a minimum definition of science. If you defined science as what scientists do, you have just achieved nothing. It's a circular definition.

If all science was based in a circular consensus, it would be in very fragile grounds indeed. The point is that it is not. You can construct a definition of science on very rational grounds. The key to that is concepts like consistency, probability theory and Bayesian inference.

Science is a process of gradually incorporating information into models to describe phenomena. It doesn't matter if it's done by "scientists" or by "artists" as long as it is done correctly. By 'done correctly' I mean that information should be incorporated in the model in an unbiased way and that the final model NEEDS to be consistent. What we identify as 'truth' and 'understanding' are very subtle things to define, but the closest we can get to them is consistency. That is the key to science and consistency can hardly be achieved by consensus as thousands of years of politics has showed to us.

So, if one asks me if "science is what scientists do", my answer is a definitely NO. What scientists do is only science when they do it right and 'right' here CAN be mathematically defined. I already wrote a bit about that here:

http://arxiv.org/abs/1008.1635

That document has many missing links, but I am preparing a larger one as I said. I will post things here as I write it.

Should you believe me or Sabine? None. This is not a question of consensus. As in science, you have to search for information and update your opinion accordingly.






2 comments:

  1. The Bayesian approach is rather typical (check Solomonov's theory of inference) yet it is not the sole root of problems. Most of all we do not yet possess a unique Science, rather a set of many and often disjoint sciences as well as methodologies. Try the below link for instance
    http://www.theguardian.com/environment/earth-insight/2014/jun/12/pentagon-mass-civil-breakdown
    What was understood as scientific activity back at the 17th or 18th century was much less fragmented and specialized then it is today which allowed for certain interactions with philosophy often fruitfull. Today's scientist is more of a "technician" and much less of a philosopher although we have kept the ceremonial title of "Doctors of Philosophy". In any case, the right balance cannot be set into a purely abstract frame, entirely cleansed from the intricacies of real life including the scientist consciousness. It is I think one of the reasons the old question first posed by Wiener remains topical: Is there a human science of human use? This enters a totality of very many parameters to admit a definite answer beyond history and social evolutions.

    ReplyDelete
  2. The main point that most people find difficult to understand is that one can formalise science by focusing on what is the main objective of it not in terms of politics or social interactions, but in terms of what kind of information one is interested in obtaining from observed phenomena. Probabilistic inference is just part of the story, not the full one. Solomonoff's contribution is one of the pieces of the full theory, but there is more to it.

    The fact that science is composed by a variety of activities and methodologies is not evidence against the existence of a fundamental characterisation of science. On the contrary, a carefull analysis of all of them clearly shows a series of common features. It is in this intersection that science is found, not in the particular details of how it is carried out.

    The main confusion, especially by social scientists, is mixing the "noiseless" concept of science with how people try carry it on and use it. Any knowledge can be used for a series of objectives and that is an important area of study, but it is not part of the definition of science. You can use nuclear physics to create a bomb. That's a military application. The process of acquiring the knowledge to do that is science, but using it to win a war is politics.

    The fact that what was considered science in the past being different of what it is now is in no way proof that science cannot be formalised. It just reflects the fact that science is not religion and its concepts can and should change as more information is acquired, just like it should be. If it wasn't like that, then something would be wrong as nothing guarantees that some of the principles we are using to do science today are indeed the correct ones. There is no contradiction here as long as you understand that those principles are self-referential.

    If most of today's scientist are more like technicians that calculate more than think, well that is a sign that maybe they should be called computers instead of scientists. To be fair, that is actually a requirement of the market. But contrary to what most people believe, the Market is not an all-powerful god that is always right and has the right to say what you are or not, although it can force you to do things. If a scientist has today to write grants and doing marketing for her research, that does not mean that it now is part of science. That's what governments and industries would love us to believe. You can still do science without all that, although you will have to find a different way to find money to keep you alive. The fact that all football players date supermodels does not mean that this is now part of the game.

    I completely disagree that science cannot be formalised and will continue to do so as long as there is no no-go theorem showing that. Of course, I wouldn't believe it is possible if other people did not have carried out works that complete the pieces of the puzzle like many philosopher and scientists like Popper and Jaynes among many others.

    The relationship of science with those who try to carry it on is important and surely cannot be separated as a human activity, but what can be formalised as science is the correct way to infer how nature works by means of unbiased consideration of acquired information. That can be formalised and the elements for it have been there for quite a long time.

    ReplyDelete