Matjaž Leonardis: Can probability really help us go beyond deductive logic?

Abstract: Deductive logic, as a model of thought and reasoning, is clearly not adequate for explaining science and the growth of human knowledge. It is often thought that extending deductive logic with probability moves us closer to a picture that is. It seems to explain the fact that evidence affects how we feel about theories that aren't implied or refuted by it and seems ultimately to justify inductive inferences once the evidence has pushed the credence of a theory beyond all reasonable doubt. In 1983 Popper and Miller published a theorem that calls this picture into question. The talk will explain this theorem and show that, on probabilistic grounds, drawing inductive inferences is always more justified before we see the evidence that is to serve as their premise than after we do.