Bayesian Rationality

I love rationality. But what is it? I have attempted to characterize it on numerous occasions, and indeed question whether it is an “it” to begin with. I view all concepts as polysemous, meaning that they have several meanings. So of course that applies to rationality too. Nevertheless there are some broad divisions that add helpful specificity to an otherwise vague or ambiguous concept without reducing it to ideological purity. 

So as regards rationality, to begin with there is deduction versus induction. Deduction establishes standards for proving things to be true. The way it does this is by stipulating rules of logic, which guarantee that if the premises of one’s argument are true, then one’s conclusion will also be true. Note, however, that this characterization is hypothetical. It does not establish that the premises of any argument are true. And so, rationality in the form of deduction can never prove anything, that is, if proof is taken to mean that a conclusion is true. Only if one could somehow also show one’s premises to be true could one rest assured that one’s conclusion is true. But logic cannot establish this, if only because one would enter an infinite regress if one then attempted to prove the premises of one’s argument to be true.[1] 

Meanwhile induction makes no claim to be proving anything. Induction is most easily understood as any attempt to defend a claim that is not a deductive attempt; in other words, there is no presumption that even if the premises were true, the conclusion must also be true. Induction does, however, assume that one can provide rational support for a claim; but now of course the rationality in question must be of a different sort from the deductive variety. In one way it is a more modest sort, since it does not aspire to proving anything; while in another way it is a stronger sort because it is not content to remain in the realm of the hypothetical but wants to roll up its shirtsleeves to tackle questions in the real world. 

But is there a positive way to characterize induction and not just negatively as not deduction? One candidate is Bayesian inference. The idea here is that one is being rational if one bases one’s conclusions on evidence, and – this is crucial – one is always open to altering one’s conclusion in the light of additional evidence. Thus there is a kind of indefinitely extended procedure for assessing a hypothesis. One begins with a given item or items of evidence and what is called a prior conclusion (or belief). I think of this prior conclusion as arising spontaneously in light of exposure to the evidence.[2] Should one subsequently become aware of additional evidence, one’s cognitive response will then remain the same or change. Here again I think of this as a spontaneous process. However, I would want the procedure to contain a further inherent element, which is that one will, to a greater or lesser degree and more or less continuously, not only be alert to new evidence but also actively seek it out. This could take the form of research, dialogue, and even just reflection.[3] 

A certain attitude would therefore attend being rational in the Bayesian sense. There is no room here for egotism: It does not matter whether one’s (“prior”) view is correct, since, in a way, truth has been put aside as a worthy or even possible goal to begin with. One is only concerned to have rational beliefs, either for their own sake or for the sake of more effective action in the world. One may therefore even be delighted to discover that one’s prior succumbs to alteration or elimination, since then one presumes to be in a better position vis-à-vis rationality or effectiveness.

[1] In a way, then, deduction is like the atheist’s God: There is the paradox that deduction is all about proving things to be true, and yet it can never do this. Similarly the very concept of God rules out that God exists, according to the atheist who argues that God is (by definition as it were) all-good, all-powerful, and all-knowing, but the world contains evil, so there is no such God. But to heap yet another irony on top of these, this argument for atheism is itself deductive; therefore it can never prove its own conclusion!
[2] Alternatively, the prior can be thought of as one’s belief prior to exposure to evidence. But it seems to me that any prior will have been the result, however implicitly, of exposure to one’s whole previous experience (in interaction with one’s nature, present circumstances, and other contingent factors); in other words, there is no state of epistemic innocence. Thus any introduction of evidence will be of additional or new evidence.
[3] In this way Bayesian inference turns out to be identical to the ethics of desirism.

Popular posts from this blog

Who Is more likely to be a psychopath: the rational moralist or the emotional amoralist?

  In the Afterword to the 2011 Edition of his book The Expanding Circle (Princeton University Press), Peter Singer makes the extraordinary admission that he has seen the light about the problematic nature of objective morality. Nevertheless he ends by cleaving to it, after providing an extremely weak argument, namely that even though a moral reason is not inherently motivating, we will nevertheless, at least in certain circumstance\s, find a moral reason to be compelling, that is, rational. To me this argument sounds almost (?) circular.  But what is even more salient for me is Singer’s conception of the rational as something that is quite distinct from what he calls the “emotional.” “[W]hether a belief gives us reasons to do something is a normative question, and whether it motivates us to do it is a psychological question.” “[I]f we can accept the idea of objective moral truths, we do have an alternative to reliance on everyday moral intuitions that, according to the bes...
Read more

The Enthusiast

I have been as ardent an amoralist as previously I was an ardent moralist. The common denominator is, of course, that I am ardent – an enthusiast. This can have amusing consequences. For example, when I was working on my doctoral dissertation, which was a defense of dispassion as a virtue, one of my advisors drolly expressed his appreciation of how passionately I defended dispassion. [1] But my general enthusiasm can also make a mess of things; and, most relevantly to a/moralism, it can, and does, create both personal and theoretical problems.  On the personal side, my general enthusiasm makes me a divided self. For as much as I am “intellectually” or “theoretically” convinced that moralism is not only based on falsity (namely, the belief in objective values, and in particular, the objective values that usually fall under the rubric of morality , such as rightness and wrongness and goodness and badness) but also has net undesirable consequences (or, as an amoralist, I would say ...
Read more

Closing the Gaps

There are a couple of gaps of reasoning in my general account of desirism, which I have just thought of a way of closing. First let me remind my reader that desirism takes two forms: There are a psychological claim and an ethical recommendation. Thus, what might be called psychological desirism is the claim that moral judgments are caused by desires; or I could put it the other way and say that desires give rise to moral claims. So for example if someone wants or likes world peace, then, I claim, they will be prone to judge as wrong the violation of peace and as bad the perpetrator of such a violation. And, in the same vein, if we judge something or someone wrong or bad, I would assume that the actual cause of the judgment is some desire, whether we know it or not.  The second form of desirism is ethical desirism, or desirism proper, which is the recommendation that we forgo making, or at least ignore, judgments of right and wrong and good and bad (etc.) and focus entirely on the...
Read more