Charles Kemp and Joshua Tenenbaum have published (or have in press) a series of beautiful papers concerned with using Bayesian inference. Their recent PNAS article opens with this sentence:
Discovering the underlying structure of a set of entities is a fundamental challenge for scientists and children alike. (Kemp PNAS)
The paper is focused on showing that discoveries about structural form can be understood computationally as probabilistic inferences about the organizing principles of a dataset.
