Paper:
Peter Godfrey-Smith (2003, 2011) argues that philosophers have made a mistake by invoking causal structure to justify inductive inferences where sample size matters, such as statistical inferences. He suggests that the secure justification for these inferences proceeds via random sampling, where each member of the population of interest has a known, non-zero probability of being in the sample. In contrast, I show that there are classes of statistical inferences that are justified by causal structure of the population of interest.

Thesis:
The core of my thesis is to show how causal structure helps us justify inductive inferences. I use the mathematics of causal structure articulated through causal models obeying the Causal Markov Condition. There are three common ways of thinking about the problems facing induction: The first is Nelson Goodman's "New Riddle of Induction". The second is the problem of constraints on prior probabilities within the traditional Bayesian framework. The third is the problem of justifying statistical inferences. I will show how causal models help us to make progress on each problem. The upshot is, I think, that epistemologists should pay a lot more attention to causal structure, especially as articulated by causal models.