Friday, January 27, 2012

Baggini, "Subjective Experiences," and Suspicion

Julian Baggini has an article up at CiF Belief responding to some arguments by Mark Vernon. I think the article sums up nicely the extraordinary confusions about psychology that typically plague Baggini's discussions of religion, and they are common enough to be worth pointing out. We see the confusions early on:

Traditional arguments for the existence of God and contemporary attempts to use fine-tuning and cosmology to back up the case for his existence always strike me as kinds of games, since hardly anyone believes on the basis of these arguments at all. Rather, they gain faith some other way and the arguments are post facto defences or rationalisations, attempts to reply to the rationalist atheist on her own terms, when the reality is the rules of engagement have never been accepted as fair. So I much prefer it when people come out and say honestly that their reasons for belief are not the kinds of reasons atheists accept as admissible, and for them to then make the case for why atheists are wrong about this.

It's not difficult to show, however, that this involves an extraordinarily simplistic understanding of the psychology of belief. The most notable thing about it is that this line of thought makes no proper distinction between causal factors leading to belief and causal factors supporting or maintaining belief. Let's take a simple case. Johnny is a physicist; he believes black holes really exist. However, he orginally came to believe that black holes really exist because someone happened to mention them when he was in the sixth grade and he felt at the time that the universe was so awesome that something like black holes had to exist. As he goes on, this same feeling motivates him to study physics; and as he does so he comes into contact with the arguments of actual physicists, based on both observations and mathematical theory, for the existence of black holes. Like any reasonable person interested in black holes, he takes these seriously; they become part of the structure that supports his belief in black holes. At no point does he take black-hole denialists seriously; he does, at times, think their arguments need to be addressed, either because he is worried about how they might mislead the public, or because he thinks they really do raise important issues on their own, and he responds in such cases either by summarizing the scientific arguments for black holes or by coming up with new arguments that are appropriate. Of course, you can replace black holes with any scientific issue you please. Now, we have here a case where belief leads understanding and where, if challenged in rational discussion, the arguments put forward all postdate the origin of the belief. Baggini, if we took his argument here seriously, would be committed to saying that all the appeals to scientific argument are in this case dishonest, because they are post facto defenses or rationalizations. This is clearly nonsense, however. Merely because something wasn't originally a reason for belief doesn't mean it never becomes one. If you want to explain the support for Johnny's belief in say, his later career as a physics professor, it would be a sign of stupidity to explain it by his middle-school feeling of what the universe's awesomeness requires. By that point, while it still may motivate him in his study of black holes, the original feeling that solidified the belief is likely one of the least of the real psychological supports of his belief.

We can take a rather different kind of case. Suppose Lisa one day becomes convinced that her husband Scott is cheating on her. She can't point to anything very definite, or even articulate very well why the things she can point to make her feel so strongly about the matter; there are just a few little things here and there that could be explained in other ways, but she can't shake the feeling that the best explanation is that he's cheating on her, although she can't really give an account of events under which it definitely would be the best explanation. She confides in her mother; and as it happens, her mother thinks it's all in her head, and raises some objections. So Lisa goes out to see if she can collect information that meets her mother's objections -- still no smoking gun, but when she comes back to her mother she has additional evidence, mostly circumstantial about his patterns of behavior, and has arguments that she's pretty sure show that her mother's objections don't apply in this case. Now, if we were to say what supports Lisa's belief at this juncture, it would be nonsense to say that "really" it's just that original sense that he must be cheating on her and that she doesn't believe on the basis of her new information "at all". This is not the way belief works. When we find new supporting information or new supporting arguments, the mere fact of their being new doesn't rule out their being grounds for belief. Quite the contrary; we add new supports for our beliefs all the time.

The problem, again, is that Baggini's argument fails to recognize that there can be any other supports for belief than the original ones. This remains a problem even if you put a great deal of emphasis on the phrase 'post facto defenses and rationalizations'. Most defenses, of course, are post facto; people often find their beliefs faced with objections they had not originally considered, and they naturally defend their beliefs. But even if the arguments in question are purely rationalizations -- we are simply making up an argument to go with our belief, one to which we have no particular prior commitment -- the quality of that argument is itself determined not by its psychological origin but by rational standards. And it can happen -- indeed, it is, I think, a common occurrence -- that what starts out as a rationalization becomes viewed by the rationalizer as a good argument and an important part of the supporting structure for their belief. Trying to say that something is not a real support for a belief because it started out as a rationalization simply shows a failure to understand how reasoning actually interacts with belief.

I also find it interesting that Baggini brings out the old trope that hardly anyone on this topic believes on the basis of the arguments. This is in fact true for most topics; the overwhelming majority of people who believe black holes exist do not do so because they have rigorously thought through the physics. Most probably have no better reason than that they heard it somewhere. We cannot tell from this, however, how significant the number of people believing precisely on the basis of rigorous physical argument may be. It could be almost nonexistent, or it could be that it's a sizable minority. People certainly are on occasion persuaded by arguments. There are plenty of testimonies of people claiming that they became theists because of this or that argument, or claiming that they became atheists when they came to the conclusion that this or that argument failed. Perhaps such people are lying or deluded, but that would have to be proven. And we simply cannot assume that they are best described as "hardly anyone". Nor can we assume, if "hardly anyone" does fit, that the arguments have no important role at all.

The muddled psychologizing continues in Baggini's later discussion:

I'm afraid it's all too common for defenders of faith to start off by piling up a whole load of interesting scientific findings, only to follow up with a plethora of non sequiturs.

The question rightly asked, however, is how reliable are the various cognitive mechanisms we use for establishing different kinds of truth? And there seems to be no escaping the simple fact that subjective experience, in all its forms, is a very unreliable detector of objective reality. Despite the comfort Vernon draws from recent research, there is no escaping the fact that the vast bulk of it points in exactly the opposite direction, undermining any confidence we might feel that our intuitive judgments are effective truth-trackers.

I find it rather amusing that he castigates believers for doing something he then goes on immediately to do. But there is a more immediate problem. The question he raises, "How reliable are the various cognitive mechanisms we use for establishing different kinds of truth?" is in fact not "rightly asked" for the reason that it is not directly relevant. Reliability is a statistical measure; it has to do with probability distribution in a population of cases. But reliability, although it can be important for assessing how strong certain kinds of inferences are, is not directly relevant here, because even if the statistical performance of a "cognitive mechanism" is very poor, if it ever, even in only one in a million cases, gives a true positive, that suffices. We would certainly want to know that it is unreliable, since it would help us to determine what we would also have to do to rule out, to a reasonable degree, that it is yielding its usual false positive, but that is all. A veridical experience is a veridical experience however rarely that kind of experience is veridical. (As a side note, it may be worth pointing out that recognition of this truth was one of the major insights that made possible the foundation of probability theory. A lot of early probability theorists originally began thinking of probability in terms of the proportion between how often, in an ideal series of cases, a given inference would turn out to be right and the total series of cases. A lot of refinements and improvements beyond this needed to be discovered, but it's not a minor point.)

Possibly what's tripping Baggini up is that we use the word 'unreliable' in an equivocal way. Sometimes when we call something unreliable we mean the normative claim that it should never be trusted because it is so rarely right that using it defeats our ends. At other times we mean it is never right. At yet other times we mean it is only occasionally right. There are sense of reliability corresponding to each of these. But it is only in this third sense, the primary sense, that it can be used in Baggini's question: the scientific evidence Baggini points to does not establish the first sense, and the second sense would make the rest of his argument question-begging at best.

Let me put the point again: drawing conclusions about veridicality from reliability alone is not itself very reliable, as a general matter. And this is for the clearest of reasons, namely, that one has to do with the properties of the population and the other has to do with the properties of individual cases. There are cases and conclusions, for instance, in which one only has to establish that something is at least sometimes veridical, however unreliably so. Likewise, one may have good reason to think something veridical in a given case even if this is often illusory; mirages may happen often in the desert, but you may still have good reason to think that you aren't seeing one despite being in the desert. The two issues are simply not the same. Reliability is important for assessing the further course of inquiry; and where we have already established very high reliability or very low reliability this can be useful for determining whether we should proceed on the assumption that any given case is veridical, where other evidence is lacking. But very general considerations of reliability -- which is all we get here -- tell us almost nothing.

Even if we set this aside, though, there is no "simple fact that subjective experience, in all its forms, is a very unreliable detector of objective reality." For one thing, this claim is radically hyperbolic. Take, for instance, your subjective experience of there being a world external to you, or your subjective experience that your wife or children are real people. We know for a fact that there are such subjective experiences because there are psychological conditions in which people don't have them. We also can be quite certain that almost everyone believes that there is an external world, or that their wives or children are real people, almost entirely on the basis of these subjective experiences. One of the most important factors leading almost any one of us to believe that there is an external world or that the humaniforms we meet are real people is that these things feel right. You will search in vain, of course, for any cognitive science research that shows that your feeling of the reality of the world around you is "very unreliable" or that you are usually wrong when you think, on the basis of your subjective experience of them, that your family members are real people. And these are hardly the only things we believe on the basis of subjective experience that either have never been shown to be unreliable or that, if unreliable at all, have never been studied adequately. The whole category is just too big to make such an irresponsible claim as Baggini is making here.

What we actually find in cognitive science research is more limited. A lot of external factors that we hardly notice can affect our judgment -- smells, for instance, or words we happened to have heard. Quite a few kinds of things that we might call 'intuitive judgments' have at least a sharply restricted range of reliability. Things that we might think closely connected -- for instance, our intuitive judgments about what is probable and probability theory -- turn out to come apart a lot. Given psychological complexity, experiments have to be very sharply defined and very precisely oriented, and so any general conclusions require interpreting a whole battery of tests -- no crucial experiments in cognition research, only experiments more or less valuable for clarification. No cognitive scientist will claim that they have rigorously studied the reliability of the whole field of intuitions, cognitive mechanisms, or subjective experiences, both in precise tests and in the wild, so as to be able to determine whether they are reliable as actually used or, with any precision, how reliable. And what we really get in the research is what you'd expect: some of these things are very reliable under some conditions and very unreliable under other conditions, and in some cases the conditions for reliability are fairly narrow. I can't avoid making an analogy here. It's rather like testing animal performance at cognitive tasks. Some animals, like border collies or dolphins are whiz-bang at all sorts of cognitive tasks. Others, like zebras, come across as utter morons at most of them. Zebras, unless there's been new research I'm not aware of, perform badly at most cognitive tasks. It's hard to say how much this indicates stupidity, since zebras are notoriously obstinate and bad-tempered. But give zebras a cognitive task requiring them to distinguish elaborate patterns of black and white stripes and it's suddenly like they're geniuses, massively outperforming animals that do better at almost everything else. And so here; a "cognitive mechanism" that doesn't track truth generally may do so extraordinarily well if certain other conditions are met. Precise specification of contextual conditions is extremely important; something may be very good at "truth-tracking" here and very poor at it there. And, indeed, if we are considering the full scope of possible situations, it is very, very dangerous to talk about truth-tracking. For the same reason, whether something is "truth-tracking" in a particular domain has to be established, not extrapolated.

Baggini ends:

The modern sceptic is indeed suspicious of subjective convictions, which is not to say they dismiss them completely. The modern believer is not suspicious enough, which is perhaps why when they try to construct arguments in their defence, the convictions are left doing all the work and reason, debilitated by neglect, weakly fails to prop them up.

I don't think, however, that modern skepticism can be put so vaguely. Most skeptics I know don't have general suspicions about "subjective convictions" even if they are suspicious of large classes of them. This is, again, the point about there being too many kinds of things that fall under such a label. Moreover, it is at least as irrational to be suspicious of something for no good reason as it is to accept it for no good reason; and at least in principle they want their suspicions as well as their beliefs to be well-grounded. And the only skeptic I've ever come across who appealed to such a ridiculously broad consideration as Baggini has here is Julian Baggini. And suspicion is as much a subjective experience as anything else: it's one of the kinds of ways we feel that something might not be right.

But if we were to use the word 'suspicious' somewhat sloppily to mean 'engaging in critical examination', instead, I fully agree that most modern believers should engage in more critical examination this way. I also think that most modern skeptics should, too. Except for where genuine practical limitations (time, resources, etc.) say otherwise, you can never have too much.