Possibility, probability and propensity

I have defended at length the idea that metaphysical possibility is grounded in the causal powers of things. It just occurred to me that this view is very naturally connected to the view that objective probability is grounded in causal propensities. We can think of probability as a measure of the degree of possibility, and of possibility as an attenuated kind of probability. If we see things in this very natural way—and hopefully its naturalness isn't just due to alliteration—then we have a unified and mutually supporting story about probability and possibility. Both are grounded in causal powers, but differently. Possibility is grounded in the bare existence of causal powers. Probability is grounded in the propensities of causal powers. If we have reason to accept one view, that tends to give us reason to accept the other.

Clearly anything that's made possible by the causal powers account of possibility—let's call this "causally possible"—is possible. So the only question about the causal powers account of possibility is whether it captures all possibilities. Suppose some things are possible but not causally possible. Then we can ask about their probabilities. If probabilities are propensities, then we should say that such things have zero probability, since nothing has a propensity to produce them. And not just the kind of "numerical zero" that classical probability assigns to a sequence of infinitely many heads, but the deep kind of zero that is had by the probability that one equals two. It's plausible, though, that things with this deep zero probability just can't happen. So the propensity account of probability neatly suggests the causal powers account of possibility.

And the converse is also plausible. If all possibilities are causal possibilities, it is very natural to measure the degree of their possibility by the causal propensities.

Of course the above is very vague. It may be that particular details of how one works out a causal powers account of possibility don't sit well with the particular details of a propensity account of objective probability.