Friday, 30 July 2010

Philosophy - Science.

Predicting future sense data

It seems that there is a past, present and future. Unlike the obvious irresistible truth of there being perceptions this sense is not sure. It is perfectly consistent to believe that there is only the present instant, that all perceptions of the past in the form of memories are simply perceptions without any past for them to reflect. I can find no argument quite strong enough to cover the purpose I'm about to put it to, this will need more thought.

We have the present, we can be certain of these perceptions, (they are all we can be certain of). We have the past, we have some memories of these (the memories are present perceptions so are certain, their relation if any to past perceptions is a question for science). There is also the future, and predicting that is the task of science.


Science is the only way to get beyond perceptions. This allows us to synthesize ideas, perceptions, memories etc and predict new things.

The scientific method is simple. First, generate several theories, a theory is a model or set of ideas that allow us to make as many future predictions as we like about perceptions of a particular kind. Then observe some perceptions. The theory with the predictions that best fit the perceptions is the best theory. This is the best approximation we have to the perceptions of the future. This does not give us knowledge in the sure and certain sense of the last post. We have to extend our epistemology (study or discussion of knowledge) and add a realm for scientific theories.

We can see clearly that this domain of scientific knowledge must be some kind of continuum. Predictions of future perceptions can be fulfilled more or less well, for example, if my best theory predicts the reading from some dial will be 1.5 and the true (observed) reading is 1.4 then my knowledge of the future movement of the dial is clearly good, but not nearly as good as it would be were my best theory to predict 1.37. Note also I've picked a numerical example for convenience, we might just as well predict seeing something blue or round. All that matters is that it should be possible to tell how well our prediction has come true. But we see that the notion of “correct” in science cannot be binary. If it were then we would consider the 1.5 theory to be just as wrong as the 1.37 theory, we cannot have improved theories, only the binary right and wrong.


The fact that this is a prediction and not a postdiction is important. We can gain nothing from knowing that a postdiction has come right. For instance, we can simply come up with the theory that all our prior observations will happen, with any prediction for the future, this will always be perfectly right. So, we require at least that the prediction not be influenced by the thing it is trying to predict.

Science as I have described it rests on the assumption that there is an arrow of time, I need a solution to this problem. I have not yet proved (at least, not to my satisfaction) that we are entitled to believe that there is a past and a future. There are two solutions to this.

Can I separate the idea of a prediction from the idea of time? In order to have science we must be able to test predictions, so we must have a notion of prediction, but can I rigorously construct an idea of predicting without having a safe idea of time? Or secondly: Can I defend the idea of time? I've no idea. If anyone can help me here please comment.

The past

It's not a big point, but a clarification. History is part of science, we can have predictions of the past as opposed to postdictions. A postdiction is a statement about past sense data, a statement about the past is a statement about future sense data that are best interpreted by talking about objects that existed in the past. For instance, I may put forward the theory that there was a Roman burial site in some location. This can make predictions, say that we could in future observe documents that refer to this site, or we could observe remains when digging there. Predictions are always about future sense data, but the best explanation for these may well be the past universe.

The world

Science gives us the world. All we can know for certain is our own perceptions, we need to construct the world “out there” based on this. The most fundamental kind of theory in science is the idea of an object. I see in front of me a black oblong shape with a large irregular patch of light in the middle. I construct the theory that this perception is caused by me seeing an object. In this case, my laptop. Please note, this is a theory, we can have sure and certain knowledge of no objects. (Brain in vat, illusion, hologram etc). But we can (and almost always do find) that this is the best explanation. The idea of an object is a theory as I described it above, because it makes predictions. For instance, an object continues to exist until something dramatic happens, and the perceptions I have “of” my laptop are reasonably constant. There is also the fact that objects continue to exist in much the same way if I move, so I predict that moving to the left will produce a different but related set of perceptions in a predictable way. And this works. The theory that objects exists is an extremely good one.

This gives an outline idea of what science gives us as good and accurate theories. I'll try and hash out in a bit more detail things like the limits of science, metaphysics, heuristics behind generating good scientific theories etc next time. But for now I'd like to suggest something that I'll try and develop later. The idea we have of science commonly is that it deals with objective things, where objective means what I identified before as perceptions of sense data or of the external world. I've deliberately defined this idea of science in a more general way, I have not mentioned other people or peer review, but I have included in "things we can have scientific theories about" such perceptions as our thoughts, imaginations, memories, emotions etc. I'm going to try and suggest later that the normal processes of science, peer review, arbitration by independent experimenters etc can be shown to be fitting with my definition. But I'm going to say from the outset that this is true for physical sciences, for a more broad idea of this second rank of knowledge we must allow for psychology, logic and other ways of predicting future mental perceptions a real place in the tent of science.

This gives the best theory

Just one last thought because I like it a lot. To the objection that science may not be the only way to gain knowledge about the world. Let us first say that we can never directly experience any object or have any form of knowledge of them except by means of perceptions. Now, suppose that some other means of discovering truth exists (be it tea leaves, holy texts, marxism or whatever), and claims that it does a better job of generating knowledge than science. Now to tell us anything that we dont already know this must make predictions about future perceptions. But, the understanding we get from science is the best prediction we have out of all the theories we have tried. Why cant we then just have as a theory "the predictions of marxism (or whatever) are correct". This is a scientific theory, if the other means really can give us knowledge then it must predict perceptions, because that is all we can ever experience that we dont already know. So then the best scientific theory must logically be at least as good as (if not better than) the tea leaves. So we can conclude (unless I've messed up this "proof") that science *must* generate the most accurate predictions that it is possible to make and that the correct way to gain knowledge is through science.

No comments:

Post a Comment

Feedback always welcome.