Monthly Archives: May 2018

Probably Right

Posted on by
11

Judea Pearl says that AI needs to learn causality.  Current approaches, even the fashionable machine learning techniques, summarise and transform, but do not interpret the data fed to them. They are not really all that different from techniques that have been in use since the early days.

What is he on about? About twenty years ago, Pearl developed methods of causal analysis using Bayesian networks. These have yet to achieve the general recognition they seem to deserve (I’m afraid they’re new to me). One reason Pearl’s calculus has probably not achieved wider fame is the sheer difficulty of understanding it. A rigorous treatment involves a lot of equations that are challenging for the layman, and the rationale is not very intuitive at some points, even to those who are comfortable with the equations. The models have a prestidigitatory quality (Hey presto!) of seeming to bring solid conclusions out of nothing (but then much of Bayesian probability has a bit of that feeling for me). Pearl has now published a new book which tries to make all this accessible to the layman.

Difficult as they may be, his methods seem to have implications that are wide and deep. In science, they mean that randomised control testing is no longer the only game in town. They provide formal methods for tackling the old problem of distinguishing between correlation and causation, and they allow the quantification of probabilities in counterfactual cases. Michael Nielsen  gives a bit of a flavour of the treatment of causality if you’re interested. Does this kind of analysis provide new answers to Hume’s questions about causality?

Pearl suggests that Hume, covertly or perhaps unconsciously, had two definitions of causality; one is the good old constant conjunction we know and love (approximately, A caused B because when A happens B always happens afterwards), the other in terms of counterfactuals (we can see that if A had not happened, B would not have happened either). Pearl lines up with David Lewis in suggesting that the counterfactual route is actually the way to go, with his insights offering new formal techniques. He further thinks it’s a reasonable speculation that the brain might be structured in ways that enable it to use similar techniques, but neither this nor the details of how exactly his approach wraps up the philosophical issues is set out fully. That’s fair enough – we can’t expect him to solve everyone else’s problems as well as the pretty considerable ones he does deal with – but it would be good to see a professional philosophical treatment (maybe there is one I haven’t come across?). My hot take is that this doesn’t altogether remove the Humean difficulties; Pearl’s approach still seems to rely on our ability to frame reasonable hypotheses and make plausible assumptions, for example – but I’m far from sure.  It looks to me as if this is a subject philosophers writing about causation or counterfactuals now need to understand, rather the way philosophers writing about metaphysics really ought to understand relativity and quantum physics (as they all do, of course).

What about AI? Is he right? I think he is, up to a point. There is a large problem which has consistently blocked the way to Artificial General Intelligence, to do with the computational intractability of undefined or indefinite domains. The real world, to put it another way, is just too complicated. This problem has shown itself in several different areas in different guises. I think such matters as the Frame Problem (in its widest form), intentionality/meaning, relevance, and radical translation are all places where the same underlying problem shows up, and it is plausible to me that causality is another. In real world situations, there is always another causal story that fits the facts, and however absurd some of them may be, an algorithmic approach gets overwhelmed or fails to achieve traction.

So while people studying pragmatics have got the beast’s tail, computer scientists have got one leg, Quine had a hand on its flank, and so on. Pearl. maybe, has got its neck. What AI is missing is the underlying ability that allows human beings to deal with this stuff (IMO the human faculty of recognition). If robots had that, they would indeed be able to deal with causality and much else besides.  The frustrating possibility here is that Pearl’s grasp on the neck actually gives him a real chance to capture the beast, or in other words that his approach to counterfactuals may contain the essential clues that point to a general solution. Without a better intuitive understanding of what  he says, I can’t be sure that isn’t the case.

So I’d better read his book, as I should no doubt have done before posting, but you know…

 

No problem

Posted on by
64

The older I get, the less impressed I am by the hardy perennial of free will versus determinism. It seems to me now like one of those completely specious arguments that the Sophists supposedly used to dumbfound their dimmer clients.

One of their regulars, apparently went like this. Your dog has had pups? And it belongs to you? Then it’s a mother, and it’s yours. Ergo, it’s your mother!!!

If we can take this argument seriously enough to diagnose it, we might point out that ‘your’ is a word with several distinct uses. One is to pick out items that are your legal property; another is the wider one of picking out items that pertain to you in some other sense. We can, for example, use it to pick out the single human being that is your immediate female progenitor. So long as we are clear about these different senses, no problem arises.

Is free will like that? The argument goes something like this. Your actions were ultimately determined by the laws of physics. An action which was determined in advance is not free. Ergo, physics says none of your actions were free!!!

But there are two entirely different senses of “determined” in play here. When I ask if you had a free choice, I’m not asking a metaphysical question about whether an interruption to the causal sequence occurred. I’m asking whether you had a gun to your head, or something like that.

Now some might argue that although the two senses are distinct, the physics one over-rides the psychological one and renders it meaningless. But it doesn’t. A metaphysical interruption to the causal sequence wouldn’t give me freedom anyway; it might give me a random factor, but freedom is not random. What I want to know is, did your actions arise out of your conscious thoughts, or did external factors constrain them? That’s all. The undeniable fact that my actions are ultimately constrained by the laws of nature simply isn’t what I’m concerned with.

That constraint really is undeniable, of course; in fact we don’t really need physics. If the world is coherent at all it must be governed by laws, and those laws must determine what happens. If things happened for no reason, we could make no sense of anything. So any comprehensive world view must give us some kind of determinism. We know this well enough, because we are familiar with at least one other comprehensive theory; the view that things happen only because God wills them. This means everything is predestined, and that gives rise to just the same sort of pseudo-problems over free will. In fact, if we want we can get the same problems from logical fatalism, without appealing to either science or theology. Will I choose A tomorrow or not? Necessarily there is a truth of the matter already, so although we cannot know, my decision is already a matter of fact, and in that sense is already determined.

So fundamental determinism is rock solid; it just isn’t a problem for freedom.

Hold on, you may say; you frame this as being about external constraints, but the real question is, am I not constrained internally? Don’t my own mental processes force me to make a particular decision? There are two versions of this argument. The first says that the mere fact that mental processes operate mechanistically means there can be no freedom. I just deny that; my own conscious processes count as a source of free decisions no matter how mechanistic they may be, just so long as they’re not constrained from outside.

The second version of the argument says that while free decisions of that kind might be possible in pure theory, as an empirical matter human beings don’t have the capacity for them. No conscious processes are actually effectual; consciousness is an epiphenomenon and merely invents rationales for decisions taken in a predetermined manner elsewhere in the brain. This argument is appealing because there is, of course, lots of evidence that unconscious factors influence our decisions. But the strong claim that my conscious deliberations are always irrelevant seems wildly implausible to me. Speech acts are acts, so to truly believe this strong version of the theory I’d have to accept that what I think is irrelevant to what I say, or that I adjust my thoughts retrospectively to fit whatever just came out of my mouth (I’m not saying there aren’t some people of whom one could believe this).

Now I may also be attacked from the other side. There may be advocates of free will who say, hold on, Peter, we do actually want that special metaphysical interruption you’re throwing away so lightly. Introspect, dear boy, and notice how your decisions come from nowhere; that over and above the weighing of advantage there just is that little element of inexplicable volition.

This impression comes, I think, from the remarkable power of intentionality. We can think about anything at all, including future or even imaginary contingencies. If our actions are caused by things that haven’t happened yet, or by things that will never actually happen (think of buying insurance) that looks like a mysterious disruption of the natural order of cause and effect. But of course it isn’t really. You may have a different explanation depending on your view of intentionality; mine is briefly that it’s all about recognition. Our ability to recognise entities that extend into the future, and then recognise within them elements that don’t yet exist, gives us the ability to make plans for the future, for example, without any contradiction of causality.

I’m afraid I’ve ended up by making it sound complicated again. Let me wrap it up in time-honoured philosophical style: it all depends what you mean by “determined”…

 

Why would you even think that?

Posted on by
27

More support for the illusionist perspective in a paper from Daniel Shabasson. He agrees with Keith Frankish that phenomenal consciousness is an illusion, and (taking the metaproblematic road) offers a theory as to why so many people – the great majority, I think – find it undeniably real in spite of the problems it raises.

Shabasson’s theory rests on three principles:

  • impenetrability,
  • the infallibility illusion, and
  • the justification illusion.

Impenetrability says that we have no conscious access to the processes that produce our judgements about sensory experience. We know as a matter of optical/neurological science that our perception of colour rests on some very complex processing of the data detected by our eyes. Patches of paint or groups of pixels emitting exactly the same wavelengths of light may be perceived as quite different colours when our brains take account of the visual context, for example, but the resulting colours are just present to consciousness as facts. We have no awareness of the complex adjustments that have been made.

This point is particularly evident in the case of colour vision, where the processing done by the brain is elaborate and sometimes counter-intuitive. It’s less clear that we’re missing out on much in the way of subtle interpretive processing when we detect a poke in the eye. Generally though, I think the claim  is pretty uncontroversial, and in fact our limited access to what’s really going on has been an important part of other theories such as Scott Bakker’s Blind Brain.

Infallibilty says that we are prone to assume we cannot be wrong about certain aspects of our experience. Obviously most perceptions could be mistaken, but others, more direct, seem invulnerable to error. I may be mistaken in my belief that there is a piano on my foot, or about the fact that my toe is crushed; but surely I can’t be wrong about the fact that I am feeling pain? Although this idea has been robustly challenged, it has a strong intuitive appeal, perhaps partly out of a feeling that while we can be wrong about external stuff, mental entities are perceived directly, already present in the mind, and therefore immune from the errors that creep in during delivery of external information.

Justification is a little more subtle; the claim is that for any judgement we make, we believe there is some justification. This is not the stronger claim that there is good or adequate justification, just the view that we suppose ourselves to have some reason for thinking whatever we think.

Is that true? What if I fix my thoughts on the fourth nearest star to Earth which has only one planet orbiting it, and judge that the planet in question is smaller than Earth? If I knew more about astronomy I might have reasons for this judgement, but as matters stand, though I feel confident that the planet exists, I have no reasons for any beliefs about its size relative to Earth.

In such a case, I believe Shabasson would either point to probable justifications I have overlooked (perhaps I am making a mistaken but not irrational assumption about a correlation between size and number of planets) or more likely, simply deny that I have truly made the relevant judgement at all. I might assert that I really believe the planet is small, but I’m really only playing some hypothetical game. I think in fact, Shabasson can have what he needs for the sake of argument here pretty much by specification.

Given the three principles, various things follow. When we judge ourselves to be having a ‘reddish’ experience, we must be right, and there must be something in our mind that justifies the judgement. That thing is a quale, which must therefore exist. This follows so directly, without requiring effortful reasoning, it seems to us that we apprehend the quale directly. Furthermore, the quale must seem like something, or to put it more fully, there must be something it seems like: if there were nothing a quale were like, there would be no apparent difference between a red and a green quale; but it is of the essence that there are such differences.

What is it like? We can’t say, because in fact it doesn’t exist. Though there really are justificatory properties for our judgements about perceptions, they are not phenomenal ones; but impenetrability means we remain unaware of them. Hence the apparent ineffability of qualia. Impenetrability also gives rise to an impression that qualia are intrinsic; briefly it means that the reddish experience arrives with no other information, and in particular nothing about its relation to other things; it seems it just is. Completing the trio, qualia seem subjective because given ineffability and intrinsicality, they are only differentiable through introspection, and introspection naturally limits access to a particular single subject.

I don’t think Shabasson has the whole answer (I think, in particular, that the apparent existence of qualia has to do with the particular reality of actual experience, a quality obviously not conveyed by any theoretical account), but I think there are probably several factors that account for our belief in phenomenal experience, and he gives a very clear account of some significant ones. The use of the principle of justification seems especially interesting to me; I wonder if it might help illuminate some other quirks of human psychology.