Last year Steph White wrote a History and Philosophy of Science honours thesis on the problem of Underdetermination of theories by evidence as it applies in Bayesian and simple deductivist models of theory choice.
In science we try to find the truth, but we usually have many theories that work with the evidence we have. This means we do not know which one is true. This is bad because everyone thinks science knows what the truth is. In this long paper I explain how this problem exists in the theory of chance worked out in numbers, and show that it is the same problem as when you model science as a yes or no question.
If we think of science as a series of choices between theories then science may have forward movement to the truth. In this long paper I explain why this forward movement can happen in a good way, and will lead us to an end point of a theory about the world that is true. This would be good for science.
The problem is that science can’t always make the right theory choice. This is because we only know some of the evidence at a particular time. But if we model science as a series of choices toward this end point we have movement in this total direction. We may make bad choices and go in the wrong direction for a time, but we make it better by the end.
We could try to bring this big change down to a small level to make science seem better if we use the theory of chance where a small set of evidence is like a big set of evidence. This does not work for many reasons because of numbers. The real problem is then that science only knows some of all of the evidence.