Varieties of Possibility: How Iterated Modalities Solve a Methodological Dilemma of Simulating under Uncertainty

There are many things about the future we know for sure. It is, for instance, certain that several million children will die of hunger next year, or that humans will not be able to live on earth once the sun has become a red giant in seven billion years. Other bits of our foreknowledge are, however, not of deterministic, but of probabilistic nature. Thus, it is very improbable that an asteroid hits the earth and extinguishes all forms of advanced life within the next decade—a probability forecast established from astronomical observations as well as geological evidence for similar impacts (Napier 2008). Still, like deterministic prediction, probabilistic foreknowledge faces limitations: There are statements about the future to which we cannot reliably assign probabilities. Section 1 is devoted to defending this thesis in some detail. Where even probabilistic prediction fails, foreknowledge is (at most) possibilistic in kind; i.e. we know some future events to be possible, and some other events to be impossible. This tripartite classification of foreknowledge is nothing but Frank Knight's classic distinction of different epistemic modes (Knight 1921). And I shall adopt Knight's terminology, "certainty—risk—uncertainty", in the following. The classification raises, in particular, the methodological question how to obtain and justify possibilistic statements. What we need is what I suggest to call a "modal methodology". How does one rationally reason in the epistemic mode of uncertainty? I will argue in section 2 that this is anything but a trivial question—and that it actually gives rise to a methodological dilemma. Solving this dilemma is the main purpose of my paper.