Turing Tests and the Non-Verbal

| | Comments (12)

Alan Turing famously devised the Turing test, which was intended to test whether a machine can think. If it could show enough behavior consistent with thinking, Turing claimed that it really does think.

Turing tests have come under quite a lot of criticism for relying on the fallacious inference from something appearing to have a certain property to the conclusion that it does have that property. Turing tests take the behavior that follows from genuine thinking to be sufficient to establish that there is such thinking, even if the same behavior can be produced by a computer program. I would take the fact that it comes from a computer program to be sufficient reason to think such behavior can occur without genuine thinking.

So the usual criticism of Turing tests is that they assume thinking is occurring just because the usual behavior resulting from thinking is occurring. While I'm not interested in diminishing that objection, it occurred to be recently that Turing tests aren't just not sufficient for thinking (things that pass the test might not be thinking). They're not even necessary (things that think might fail the test). For one thing, someone who thinks might simply refuse to comply with the test and thus could fail. But more poignantly, someone with a communication-related disorder, e.g. someone with autism and dyspraxia who is completely non-verbal, simply cannot display the behavior the test is looking for. Being unable to communicate is certainly not a sign of being unable to think.

I would argue that more harm is caused by those who take passing a Turing test to be necessary for intelligent thought than is caused by those who take passing such a test to be sufficient for intelligence. We recently attended a communication seminar for parents and educators of non-verbal and mostly non-verbal children. At one session an autistic college senior was present. He can now speak in a somewhat limited manner, but he can communicate by typing on a portable device at a level that's almost certainly far beyond what most kindergarten teachers would have ever expected if they had seen his communication level in his younger years. He had no verbal language until age 12, but because his teachers taught him to type they knew that he was able to grasp much higher levels of thought than most teachers would have even speculated. At last night's session, there was a guy with Down Syndrome and autism who, as far as I could tell, can even as an adult do little more than grunt was typing out sentences that indicate a pretty high-level grasp of some pretty abstract and complex phenomena.

With a son who can't speak much more than five syllables at a time (unless he's singing or engaging in echolalic repetition of Veggie Tales or some other TV show), we've been able to see something like this firsthand. We knew in kindergarten that he was reading fairly complex words for the level of verbal behavior we normally saw, because he'd occasionally see a word and say it. (I remember him saying "banana" one time when there were no pictures of a banana, just the word.) But it's been very hard to get him to demonstrate his intelligence with writing, until this year, with his teacher and support staff working very hard with him to get him typing. Six months ago we could get him to trace over words we wrote out with a highlighter, or we could get him to point to words sometimes on a communication device, which could then pronounce them for him (but they had to be programmed in first, since he wasn't typing them). Now he's showing reading comprehension by completing "because" clauses to answer why certain characters did certain things. It makes me wonder how much he's been wanting to be able to communicate for years but unable to get his mouth or hands to do anything to show it.

The Turing defender might now say that he is able to show it, so it's not an objection to the test, but he's only now able to show it, and there's no reason to think he just started to be able to think on this level. I suspect most teachers would have assumed he couldn't handle the level of math that he's doing (basically right on second grade level) or the vocabulary and reading that he's doing (which is, as I said, at a pretty good level for demonstrating reading comprehension, better than his older brother could demonstrate at that age). He happens to have a teacher with 25 years of experience working with kids like him, who is informed about technology and methods to get kids like him communicating. Many educators encountering a kid like him might well assume low ability levels and not work to get him to communicate. In effect, they're using a reverse Turing test and concluding that someone isn't intelligent because they can't show it in the typical ways.

Enhanced by Zemanta

12 Comments

‘It makes me wonder how much he's been wanting to be able to communicate for years but unable to get his mouth or hands to do anything to show it.’

All kids get to do things they didn’t do before. They’re on a learning curve. Traits innate to the human brain may be associated with adaptive behaviour which helped ancestors survive and reproduce, but that’s a hypothesis to constrain logical space and explain how learning, e.g. language acquisition, is possible in a certain environment. You can look at a cognitive theory as an instrument for explaining and predicting behaviour, or search for biological correlates; there are ways to map brain function with fMRIs for example, but stimulus-response intervals are measured in seconds rather than years! I suspect you may have something more elusive in mind, in which case I can’t think how the question could be decided either way actually; perhaps it’s more metaphysical than scientific. (PS Does the brave boy enjoy cuddles?!)

I think you're pushing the Turing Test into something it's not. Passing a Turing Test is evidence thinking. Failing a Turing Test is not evidence of an absense of thought.

In medical terms think of 'false positives' versus 'false negatives'. A pregnancy test, for example, can read negative when the woman is really pregnant (esp. very soon after conception). Hence a positive pregnancy test almost always means pregnancy, a negative test, though, does not establish a condition of 'not pregnant'.

If you mean to be saying that no one has ever taken the absence of thinking-behavior to be a sign of something's not thinking, then the very example I gave seems to me to be a pretty clear counterexample. There are plenty of others. No one thinks a bicycle thinks, for instance, and the reason many think that is that it can't pass a Turing test.

That's hardly 'the reason' people think that a bicycle doesn't think. The more likely reason is that a bike lacks any internal complexity that would lead a person to think it could generate thought. But technically Turing simply would have to say we don't know, the bike's failure on the Turing test isn't proof that it doesn't think.

Likewise a person under sedation can't pass a Turing test either. Yet not only do people believe such a person thinks, they even can tell if the person is dreaming or not.

The argument that the bike lacks the internal structure relevant for thinking relies on the assumption that things cannot think without the internal structure of things that pass Turning tests. There's still an implict reverse Turing test going on to justify the induction.

Turing may well have not taken his test as a necessary condition for thinking (or may have been agnostic). Nevertheless, most of us do when we assume children are mentally deficient if they don't talk and don't show behavior of complex thinking, which is what I'm complaining about. The sedation case shows not that we don't do this but that we're not consistent about it.

Well such children are mentally deficient. We want children to not only think but to be able to communicate and interact with others. If I told you a bicycle is able to think you may be surprised but you probably wouldn't want to be a 'thinking bike' as opposed to a 'thinking person'.....communication likewise is a good that takes thinking to another level. Hence you are happy as your son's accomplisments increase. If these accomplishments were not in themselves a good thing you'd be indifferent to them. If they are a good thing then that means their absence would be a bad thing. That doesn't mean someone who can't accomplish a particular level isn't human, shouldn't be loved etc. But it does mean that something is wrong or deficient.

1. The complexity issue is actually a hypothesis...thinking is either caused by or travels through complex arrangements of matter. Hence we rate it highly probable that a brain or complicated computer might think but we doubt that a piece of soap thinks.

2. The person under sedation indicates that we rely on multiple tests and hypothesises.. We know the sedated brain and the sleeping brain are typically 'paused' as far as thinking goes only. We know the disruption to the brain's complexity is temporary. We don't know for sure if a sedated person 'is thinking' but we know the sedated person thinks in the sense that you know your sleeping wife 'walks' in the sense that she is an entity that is capable of walking....even if not at this particular moment.

I suppose then you can say we are talking a calculated risk in assuming the bike does not think. Maybe it does but before we fret about melting down an old bike, if inanimate matter can think then why would it matter if the bike was a bike or a melted down block of metal?

Sci-fi often supposes that thinking robots and computers would face a lot of discrimination by humans who would insist that they are 'only machines' but I don't think so. Most people are very happy to extend the idea of thought beyond themselves. How many pet owners go to 'dog whispers' because they yern to know that thir dog can think and would love to hear its opinions on things. People would probably attach themselves quickly to 'thinking machines'....at least if the machines seemed to think in a manner that was near human. Likewise people who have loved ones in a brain dead state....maybe even in a casket....will swear they see some response indicating thought even if we know they haven't. At least when we want to see it, we are more likely to see thinking when it's not there rather than the other way around.

That's where the big leap question is IMO. Would humans be able to recognize non-human thought? Turing wasn't really testing for thought but human-like intelligence.

You're confirming my point if you assume that an inability to communicate means mental deficiency. I'm not going to join the disability community's new orthodoxy, which says that disabilities are just differences and not deficiencies. That's the result of the loss of any sense of divine intention in our abilities, which I don't think a Christian can consistently adopt. So I wouldn't claim that autism involves no deficiencies (but I would want to add to the insistence that there are genuine disabilities that some features are also heightened abilities).

Is it best to think of these as mental deficiencies, though? Perhaps in a broad sense, since they involve the mind. But are most of them to do with what we usually call intelligence? I'm becoming less sure of that the more I understand it. To be sure, autism usually involves some kinds of cognitive deficits, but those mostly involve social understanding, sensory hypo/hyper-activity, sensory integration, and motor planning. Sometimes there are attention deficits.

The element that's best thought of as a mental deficiency is the difficulty with more abstract thought and the focus on the concrete, but I suspect that's not really an inability but a reflection of interest. the sensory integration issues explain the hard time autistic kids have in focusing on the work at school, and motor planning seems to be the major difficulty with my second son in getting his mouth to utter the words he's thinking. It's not really an issue of mental competence at all in the narrow sense of what we usually call intelligence, and anyone interacting with him and trying to get him to talk would probably assume that it's purely an issue of mental competence, since the usual ways of observing intelligence don't work so well.

I'm not sure thinking is either caused by or travels through complex arrangements of matter. It's certainly associated with that sort of organization of matter in all the animal cases we know about, but I'm not going to concede that that's the cause or location of thought in the human case, never mind that it's the only possible cause or location of thought. I'm committed to some kind of dualism, and I don't take God or angels to be material beings at all, and I don't doubt their thought. Even if Descartes was wrong about how dualism works (in which case Aquinas would be closer to the truth), I'd still have to make sure whatever hypothesis I come up with handles God and angelic beings.

The sedation case is usually dealt with by taking thought to be dispositional at some level. You have a belief if you're disposed to do something under certain circumstances. You have a desire if you're disposed to do something under certain circumstances. I'm not sure I'd accept an account that defines thought in such terms, but it does seem to be true of thought, and that element of the nature of thought easily allows gaps in one's ability to think without denying an ability on a more general level. It's kind of like recognizing that I have an ability to speak Finnish that isn't being realized because I've never tried to learn the language, and so I don't have the ability in the way that a native Finnish person does, but I have it in a dispositional sense in a way that an ape doesn't.

I agree with you that some people would overinterpret what they see as signs of thought and form emotional attachments to machines that have the behavior of thinking things. But that doesn't mean the scifi fear isn't based on real human tendencies. After all, we've consistently denied equal treatment to people who to all appearances are as fully thinking as the rest of us, usually with some excusing statement that the group in question isn't as advanced as the rest of us.

1. I would agree w/you regarding deficiencies. I would go in with the 'community's new orthodoxy' in the sense that differences may have their own advantages which shouldn't be ignored. Likewise as the definition of mental deficiency expands to cover things like ADHD, I start to wonder whether maybe we are just defining differences as deficiencies. Maybe the person who doesn't like to sit in a classroom is perfectly normal and would have seemed so a thousand years ago but just doesn't fit in with our modern society....maybe we need more alternatives to classrooms rather than drugs to treat 'alternative' people.

2. Its been a common argument in pro-life circles that notorious cases of past oppression have been by groups that 'denied personhood' to 'out groups'....examples being Nazis against Jews and slavery supporters against blacks. I think, though, this gets it backwards. They embraced the personhood of their victims and rejected equality. At some points it was a crime to teach a slave to read in the US. It's never been a crime to try to teach a cow to read. Slave holders didn't think blacks weren't people but cows were, they knew exactly what they were doing and knew that because blacks were people the tables could be turned on them someday in a way that cows would never turn the tables on humans....hence a cow who could read would be a hoot, a slave who could would be a dire threat.

In general I think we are more likely to overread personhood than underread it. Babies will instictively respond to a smiling face...even a non-human one. Hence cartoons and toys work by attaching our quality to attach to anything that seems remotely like a person.

As for the question of whether thought depends on matter only or some type of duelism....I don't really think the question is all that important. At least I dont' think it has the weight in the atheism.v.theism debate that both sides like to think it does. I'll note that if humans do end up oppressing intelligent robots, it will probably be because they are too caught up in the matter.v.deulism debate.

ADHD is way overdiagnosed, but some kids are helped tremendously by medication. In the case of my oldest son, it takes about ten attempts to get his attention just to get him to write his name. He can focus just fine when the activity is intrinsically rewarding to him, but his brain can't get the deferred-gratification reward system going, and without the medication he won't be able to develop good habits of long-term planning and doing unpleasant things now to reap good rewards later. It's not just an issue of calming kids down so we don't have to deal with their high activity. That's a stupid reason to medicate a kid, perhaps even an immoral one, unless it involves such extreme impulsivity that his own and others' safety and well-being are at stake (as in the case of my second son). But if normal moral and rational development is at stake because of an inability for the unstimulated hypothalamus to develop this kind of focus to do important things that will benefit his life greatly in the long run, I think it's immoral not to medicate.

I agree with you that there never really was any widespread belief that blacks or Jews are not persons (and the legal fiction of the fractional person in the U.S. Constitution is no exception to that, despite its constant use to try to support some sort of partial personhood view having existed at the time). But that's neither the issue I was talking about nor the issue that pro-lifers who make the argument you're referring to really mean.

What pro-lifers are concerned with is full moral status, however that is defined or whatever properties it's supposed to be grounded in. The last few decades have seen philosophers treating personhood as the proxy for moral status, but you can ignore personhood entirely and just deal with the more fundamental issue of moral status. The real issue in their argument is that moral status wasn't being ascribed to the groups they're pointing to historically, and the same is true of the fetus today. I don't think the argument has to rely on false history to make the point they want to make.

What I was talking about also wasn't really personhood but intelligence, which can come in degrees and might in theory be had by some things that aren't persons. It certainly can be had in very low degrees by some persons. The reverse Turing test that I see in operation is assigning a status of mental retardation to someone who may be very intelligent in the traditional sense of intelligence (by which I mean what an IQ test tests in the cases when it can accurately show the skills that it's trying to test for, as opposed to other kinds of intelligence like emotional intelligence or whatever else people are going to call a kind of intelligence, and I do think there are such ways of comparing various cognitive abilities). Someone could be the sort of person who would do very well on an IQ test if the person could be tested, but other factors prevent the person from being tested. The same factors prevent the kind of behavior a Turing test looks for, and in fact it's a kind of Turing test that they use to evaluate how intelligent someone is when selecting educational plans in special education. It's testing for behavior consistent with intelligence. If we conclude after not seeing behavior of a certain sort that the person is of low intelligence, when the obstacle to displaying the behavior in question is a problematic pathway between the speech center in the brain and the muscles that move the mouth, then the reverse Turing test gives the wrong result. That's the kind of thing I'm talking about, not the kind of claim that people who can't talk aren't really people.

Minor quibble, the Constitution did NOT define slaves as 'fractional persons'. The south wanted slaves counted fully for purposes of representation. The north did not on the grounds that slaves were not allowed to vote by the south. The 3/5 clause was a compromise between the two points, note that the pro-slavery position wanted slaves counted fully.

You're preaching to the choir on this one. That was in fact my reason for saying it's not a genuine exception but just a legal fiction.

To EO: I'm not saying there's no physical evidence of someone wanting to talk in such a case. I'm saying the kind of evidence a Turing test would produce isn't going to show the intelligence he has. Turing tests don't involve using an fMRI machine.

He actually loves cuddling, at least with people he's familiar with. He has much less of a deficit in social interaction than most autistic people do. His deficit is mostly in communication and somewhat in cognitive processing due to sensory issues. His brother is much less social but much better at communicating.

Leave a comment

Contact

    The Parablemen are: , , and .

Archives

Archives

Books I'm Reading

Fiction I've Finished Recently

Non-Fiction I've Finished Recently

Books I've Been Referring To