I'm delighted to announce that “The reliable route from nonmoral evidence to moral conclusions” has been accepted for publication in Erkenntnis. Here’s how it begins:
"There is a reliable route from nonmoral evidence to moral conclusions. Progress through its three stages relies fundamentally on inductive inference. First, we divide up the psychological processes generating belief so that their reliability in generating true belief is statistically predictable. Second, we measure their reliability – the proportion of true beliefs they generate. Third, we infer probabilities of truth for moral propositions from the reliability of the processes generating belief in them. The three parts of this paper map out the three stages of the reliable route."
Part 1 discusses what processes of belief-formation are, how to individuate them, and how to use their reliability in reasoning. Processes are things like visual perception and wishful thinking. Since we know that visual perception is more reliable than wishful thinking, we change beliefs that we think were formed by wishful thinking, but we generally keep beliefs that we think were formed by visual perception. James Beebe and Jack Lyons’ work on the generality problem for reliabilism helps one imagine such belief-forming categories defined with scientific precision, perhaps with equations predicting how likely it is that beliefs caused through various processes under various environmental circumstances are true.
Part 2 discusses how to determine the reliability with which moral beliefs are formed. We can’t assume any moral truths in determining the reliability of the processes, or else the reliable route would be circular. I offer two ways to infer the reliability with which nonmoral beliefs are formed, beginning from entirely nonmoral information. First, reliability can be inductively inferred from that of similarly generated nonmoral beliefs. (Most metaethical theories have the same or similar processes generating moral and some nonmoral beliefs, often a form of intuition or perception.) Second, contradictory moral beliefs push processes generating them towards unreliability, regardless of which belief is true.
Part 3 begins as we've derived things like this in the first two parts:
Process P, which generates belief in moral proposition M, has truth ratio T.
(T is the proportion of true beliefs generated by P, with parameters filled in)
We can now revise beliefs in M accordingly. There’s some kind of probabilistic Moore’s Paradox in saying “M, and there’s an 0.1 probability that M.” If we’ve always been confident in M, but P alone generates belief in M and it has truth ratio 0.1, P is essentially a cognitive bias on moral judgment and we should give up belief in M. Similarly, if we discover that one of our moral beliefs is formed by processes especially reliable in generating nonmoral belief, we should become more confident in it. If we gave it up because it conflicted with another moral belief that we now discover was unreliably formed, we should return to it.
Inductive inference takes us all the way to moral truth. We empirically discover the reliability with which we generate various nonmoral beliefs, inductively apply these observations to similar cases in which moral beliefs are generated, and infer probabilities of truth for similarly generated moral beliefs. Using the reliable route, we cross Hume's famous gap between is and ought, on the power of induction alone.
