Anti-Mind, pt. 2 of 2

The world is everything that is the case

Originally uploaded by Pete Mandik.

Does it take a mind to detect a mind? If there could be a principled answer to this question the implications would be huge for the philosophy and science of mind.

Consider that so much of science depends on the unintelligent detection of unintelligents. Hydrogen samples are not particularly intelligent. Further, mechanisms capable of detecting the presence of hydrogen need not themselves be intelligent.

Maybe part of being a natural kind is that the unintelligent detection of instances of that kind is possible. Jerry Fodor has suggested that non-natural kinds like crumpled shirts or doorknobs can only be detected by minds. You have to be the sort of thing that knows a bunch of stuff in order to “light up” in the presence of a door knob.

In the Sterling short story “Swarm” excerpted in my previous post, the Nest is this asteroid that is mostly just a big super-organism that wanders the universe and whenever it is “invaded” it assimilates the invaders. Most of the various diverse species in the asteroid were once representatives of vast space-faring technological cultures that, when they encountered the Nest, got taken over and reduced to unintelligent animals and integrated into the Nest ecology inside of the asteroid. Swarm is an intelligent organism activated under certain instances for the protection of the Nest. Swarm explains how ultimately useless intelligence and consciousness is and suggests that the Nest is entirely unintelligent, and that the Nest grows a new Swarm whenever an intelligent invader needs to be dealt with. Once the intelligent invader is dealt with (rendered into a dumb slave animal) then Swarm self-destructs being no longer needed.

It occured to me that Swarm was to minds what antibodies are to germs, so I coined “anti-mind”. It also occurred to me that if Swarm was right that prior to the activation of Swarm, the Nest group organism was truly non-cognitive, then whatever mechanism that activates the growth of a new Swarm must itself be an unintelligent mechanism. So, the idea of an anti-mind is the idea of a thing that is not a mind but is capable of detecting minds. But this leads to what strikes me as some pretty interesting philosophical questions: Is there any way a dumb mechanism can detect the presence of intelligence? Can an unconscious mechanism detect the presence of consciousness?

If Dennett is right, intentional systems are detectable only from the intentional stance, which I take to entail that only minds can detect minds. If a lot of qualia-freaks are right, the only way to detect the presence of qualia is to have some yourself, and thus only consciousness can detect consciousness.

If these remarks are correct, the implications for science fiction are obvious: the “anti-mind” in the Sterling story is impossible. But enough about fiction: what about science? If the impossibility of unintelligent detection entails that the kinds that are intelligently detected are non-natural, then is a full-blown science of such kinds thereby doomed?

12 Responses to “Anti-Mind, pt. 2 of 2”

  1. Tad says:

    Cool post, Pete.

    What about the modularized intentionality detector that many developmental psychologists claim comes on-line very early in development? Though it’s part of a developing mind, it could be argued that the module itself is unintelligent, and triggered by relatively low-level behavioral cues (think of the Heider & Simmel result - geometric shapes moving in certain patterns automatically trigger mental state attribution).

    Do your concluding questions suggest that by Fodor’s own criterion, minds can’t be natural kinds? That’s not his view, so I suppose he’s committed to the detectablity of minds by non-minds. If Dennett’s view is right - that minds can only be detected by minds, does an infinite regress loom, i.e., something only counts as a mind if detected by something that only counts as a mind if detected by…?

    I’m a little skeptical that you can only have a science of natural kinds, where these are things that make a difference to non-minds. I think you can have a science of Ford Tauruses - there is projectible, quantifiable information you can learn about Ford Tauruses. Computer science is a much better example. Surely there are counterfactual supporting generalizations about computation, and isn’t this enough for science? Yet aren’t computers human artifacts, and therefore dependent on minds? Perhaps Fodor would argue that computation can make a difference to non-minds - I guess he’d have to argue that, given that the computational theory of mind states that computation makes a difference to behavior.

    It occurs to me that there are a couple of issues conflated here. It could be that something can’t exist without a mind, e.g., fossil fuel technology, yet, once it exists, have effects and hence be detectable by non-minds, e.g., global warming. So mind can play a constitutive role in the existence of something, without being the only means for detecting it.

    One more question: when you ask whether minds can be detected by non-minds, do you mean qua minds? It could be that there is some physical signature that all minds have in common, though merely *in extension*, that could affect non-minds. Consider nonequlibrium thermodynamics. Churchland (1982 ‘Is Thinker a Natural Kind?’) argues that living and intelligent beings actually have an important physical property in common: they’re non-equilibrium thermodynamic systems. That is, they’re part of an energy gradient, absorbing energy while giving off heat. Given the laws of physics, one could see why minds might all give off this physical signature. But this might not be a feature of minds qua minds, because we could imagine worlds with different laws where minds don’t have it.

  2. Pete Mandik says:

    Thanks, Tad.

    I don’t have strong opinions about this stuff or a particularly worked out view, but was hoping that thinking out loud about this stuff would inspire some further thought. I think you raise some pretty interesting considerations, so my dreams came true.

    A lot of this stuff is going to be hard to pin down because the relevant notions are all pretty underspecified: detection, intelligence, natural kinds etc. However, here are some of my hunches about this stuff.

    1. I’m inclined to think that it is a criterion on natural kinds that they be detectable untintelligently.
    2. Whatever the relevant notion of detection is, it better be something involving detecting so-and-so as such. This is, of course, hard to pin down, but maybe something like counterfactually supporting laws is relevant. Detecting greenhouse gases in an atmosphere, though caused by humans, probably doesn’t count as detecting humans. More on this below.

    Regarding some of your specific remarks:

    Re: the intentionality module, I wonder just how unintelligent it is. There ought to be ways of operationalizing the difficulty of various identification tasks. For example, certain discriminations can’t be learned by feed-forward networks. This module is probably more complex than that. But I don’t know much about measuring the complexities of various problems. Also, how far do we want to go in saying what the module detects. Would setting it off count as passing the Turing test?

    Re: Dennett’s regress. That’s a charge often raised against him, and I’m not clear on what his canonical response is. Something to do with cranes and skyhooks.

    Re: Ford Tauruses. Really? qua Ford Tauruses? That strikes me as odd, but perhaps I just haven’t thought about it enough yet. Any particular examples?

    Re: computers. That’s a tough one. I’m inclined to think of computer science as not really a science, but like applied math or something. It doesn’t make predictions so much as work out entailments about formal systems. However, I fully admit that I barely know what I’m talking about here.

    Re: the conflation question. I think you raise a really interesting point there. Fossil fuel technology is indeed an artifact. It’s products though are chemical compounds and thus natural kinds. They may not be especially “naturally occurring” natural kinds, and thus a not bad indicator of people. This is a tricky issue. Maybe part of the problem is that “natural kind” is a concept that needs some further work.

  3. Tad says:

    Pete -

    With regard to Dennett, I think there is a better response. Patterns are real whether anyone detects them or not. I think the way to think about his proposal is as claiming that some patterns are more interesting/useful than others. So the pattern tracked from the geocentric stance is just as real as the pattern tracked from the Galilean stance, only the second one is much more useful, subsuming most of the usefulness of the first one (though not all - navigation on the open waters often relies on geocentric assumptions).

    You can define sets of systems in terms of the patterns they generate. Intentional systems are those that generate patterns trackable from the intentional stance, whether or not any mind notices them by adopting the intentional stance. How interesting/useful are these patterns, and how useful is it to group all such systems into one group? That’s the interesting question. I’m not sure why we care about whether this group is a natural kind, or real. All patterns are real - that is any phenomenon from which we can glean useful, projectible information, is worthy of study, and theoretical systematization. But some patterns are more robust and useful and interesting than others. So, the question becomes, is there any reason to think intentional patterns are more robust or useful than say patterns generated by Ford Tauruses qua Ford Tauruses? (BTW, there are plenty of ceteris paribus laws about Ford Tauruses. E.g., ceteris paribus, Tauruses made before ‘92 last longer, etc.)

    I think yes. Dennett speaks of intentional patterns as those that are there whenever there is a selection process present. I think intentional systems and the distinctive patterns they generate have a kind of deep, unifying explanation in such terms, that makes them as close to a natural kind as anything. There is a kind of ‘edge of chaos’ dynamic to certain physical systems with certain thermodynamic properties that gives rise to selection for structures stable enough to make a difference to the world, yet sensitive enough to variation to invite selection, and hence intentional interpretation. So the patterns such systems generate are relatively robust and interesting. Why isn’t that enough? Why privilege some kinds/patterns as natural? Why not treat pattern, or information as ontologically basic, and then look at which patterns are particularly robust and interesting, and why?

    Anyways, sorry if I’m hijacking this a little. Didn’t mean to go off on such a tangent! If you think living kinds are natural, and you agree with Dennett, then you should think minds are natural, since living kinds and minds are all intentional systems of varying sensitivity to the environment. I think living systems are detectable unintelligently - since living systems detect each other, and not all of them are intelligent. But whether something that’s not an intentional system in Dennett’s extended sense could detect anything is doubtful. Detection sounds like an intentional act.

    Perhaps by ‘detection’ you mean nomic connection. So the question is, can intentional (including biological and psychological) kinds, qua intentional kinds, enter into nomic connections with non-intentional kinds? What would you say if all and only intentional systems had some higher-order physical property in common in the actual world, and that physical property entered into nomic connections with other physical (unintelligent) properties? I take it swarm could then use this information to detect intentional systems. But he’s not detecting them qua intentional, but qua the higher-order physical property. But who cares? Shouldn’t we just worry about the actual world, or nomically possible worlds? And maybe in all nomically possible worlds, all intentional systems do have some higher-order physical property in common.

  4. Nick Treanor says:

    a very interesting question, and certainly worth thinking about.

    Here are a couple of thoughts, dashed off before morning coffee.

    First, I wonder how much is meant by ‘detection’ (in particular, I wonder whether it is factive). I’m thinking of this especially in terms of other minds — it is unclear whether people are good mind-detectors at all (witness the debate about animal consciousness). For this reason, it is unclear whether even a mind can detect a mind (other than, perhaps, itself). This problem arises I think even if ‘detect’ is weakened to require only reliability. (Suppose no animals are conscious, and that most people think most of them are. Then we are not even reliable at detecting consciousness….most of the time we are mistaken.)

    Second, and somewhat related, presumably other people have minds and our ‘mind detector’ lights up when we are around them. This could be because there is a contingent correlation between gross bodily movement of a certain sort and having a mind. So we ‘light up’ in the presence of gross bodily movement of that kind. That’s what triggers the detection module. But there may well be worlds (even with the same laws) where gross bodily movement of that kind doesn’t correlate with minds. We don’t have to imagine zombies…just imagine bodies shaped like ours, stuffed with straw, that are something like puppets, controlled by some kind of hidden mechanism. My guess is that our ‘mind detector’ would light up just like usual (because it is responding to something like gross body movement, not the presence or absence of mind).

    Third, if minds supervene on the physical, then wouldn’t the causal exclusion argument suggest that (assuming detections of minds is possible at all), that physical systems can after all detect minds? The idea here is just that if I can detect, for instance, that you have a mind, then surely my brain is what did the detecting. If this is right, then it may still be that ‘only minds can detect minds’, but the punch would be taken out of this because it would mean only something like ‘only physical systems of a certain kind can detect minds’, which isn’t much to swallow. (Lots of things can only be detected by physical systems of a certain kind, but nothing seems to follow from this regarding whether what is detected is a natural kind, does it?)

  5. Pete Mandik says:

    Hi Tad,

    I like the nomological characterization of detection you offer in the last paragraph, and I think the questions you raise there concerning whether the detection of the intentional would have to be qua intentional are pretty important to the natural kind questions.

    I think a lot of this can be discussed in terms of what’s really going on with Fords, so I’m going to press you on that. I’m still skeptical that Fords enter into any laws qua Fords. I would think that they, e.g. last longer, not because of their being Fords but because of their, e.g., having more resiliant cam bearings than non-Fords which is due to their cam bearings being made out of titatnium. In otherwords, what makes them a Ford is a bunch of historical stuff that doesn’t much matter to their longevity. And what matters to their longevity is a bunch of chemical/metalurgical stuff that doesn’t much matter to their being Fords.

    I don’t know if I have any actual arguments, but I like the view that what makes a kind natural is that it enters into laws qua that kind with things other than intentional systems that identify them as such.

    Regarding the “higher-order physical property” of minds you mention in your last paragraph, I would say that, yes, if mind had that in all nomically possible worlds, then the unintelligent detection of that property would save minds from being unnatural kinds.

  6. Pete Mandik says:

    Thanks, Nick!

    Re: how good we are. I know this isn’t particularly precise, but I’d say: we’re pretty darn good. At least, we detect minds about as well as we detect other things. Set us loose at a party and we know well enough not to attempt a conversation with the sofa or the punch-bowl. Regarding natural kinds, take water: we detect it well enough to stay hydrated for a good long life. And then we got clever and figured out how to build machines that can detect the hell out of water. We’ve yet to get sufficiently clever about minds to tell whether, e.g., earthworms have ‘em. I think this casts a pretty dark shadow on whether we really have a scientific grasp on minds.

    Re: your last paragraph, I’m not sure I follow the punch-removing powers of the exclusion argument here. Suppose that something could be detected only by physical systems that instantiate minds. That strikes me as a pretty clear indication of being a non-natural kind. I say this because I’ve gathered up several prototypical instances of non-natural kinds (obnoxious shirts, drawings of fish, cats named “Larry”) and it strikes me that the only detection supporting causal interactions that they enter into qua those kinds are the ones they enter into with (physical systems that instantiate) minds.

  7. Nick says:

    Pete, I had forgotten that I’d commented on this thread. (It must have been the fact I hadn’t yet had coffee!) So this reply comes a bit late.

    I don’t see why you think that being detectable only by a physical system that instantiates a mind is a sign of being non-natural. The argument couldn’t be:

    (i) shirts, fish drawings, etc., are non-natural
    (ii) they can only be detected by minds
    Therefore (?) things that only minds can detect are non-natural.

    This is obviously a caricature of your argument, and isn’t what you had in mind. Help me see what you take the argument to be.


    As a separate point, I take it that you are committed to the following claim, where “structural complexity P” is some realization base of mentality:

    if x is a property that can be detected only by a physical system with structural complexity P, x is not a natural kind

    To me this seems like a very odd claim. Is there any reason to think that if something is a natural kind, it can be detected by ’simple’ physical systems? And if not, why doesn’t P get to be one of the respectable, physically complex detection systems?

  8. Pete Mandik says:


    I think this merits much further thought, and I hope to blog on this again soon. In the meantime, here’s a bunch of thoughts to hopefully convey where I’m coming from.

    1. Natural kinds are a subset of the real kinds. So, there really is such a kind hoaks: instances of which are mereological fusions of one oak tree and the nearest horse. While that’s a way of carving nature, it misses the joints. Hoaks form a real but non-natural kind. Oaks and horses, in contrast, are natural.

    2. Natural kinds are the sort of thing such that quantifying over them and their instances is a good way to build a science.

    3. Science building is supported when the things referred to enter into lots of nomic relations. How do you know if you’ve carved nature at the joints? If your carving yeilds surprising predicitons, for example.

    4. Consider pens. An incredibly diverse number of objects count as pens. Writing with ink is not something that all pens do: broken pens and empty pens are still pens. Pens don’t enter into many law-like interactions qua pen. Except for one: humans know ‘em when they see ‘em. There won’t be a science of pens since there’s not much to say about them beyond enumerating which is a pen and which is not. Same goes for what counts as art, what counts as jazz, and what counts as porn. If there’s nothing else to say beyond “we know it when we see it”, then that’s a good sign you arent’ carving the world at its joints. You are just carving the world.

    5. The only thing that can indicate the presence of jazz is a sufficiently informed person who can judge “yeah, that’s jazz”. In contrast, lots of things beyond suitably informed persons can indicate the presence of e coli, of hydro sulpheric acid, and of quasars. Instances of these latter kinds enter into nomic relations with lots of other stuff in the universe besides the people who talk about them.

    I’ll stop there for now. Maybe something connects with your worries. Or maybe I’m just continuing to beg the crucial questions.

  9. Tad says:

    Pete -

    Thanks for such a clear statement of the view. Only one question: are you receptive to the suggestion that natural kinds differ from non-natural kinds in degree, not in kind? The boundary you suggest sounds vague: “Science building is supported when the things referred to enter into lots of nomic relations” - when is enough enough? It seems like it would be more perspicuous to say that some kinds are more natural than others, instead of claiming there to be a strict, precise distinction b/w natural and non-natural kinds.

    One more problem. Unless you’re willing to countenance the view that only physical kinds are natural, there may be a problem in requiring that kinds must enter into nomic relations qua the kinds that they are, in order to count as natural. Here is why - arguably, all but physical laws are ceteris paribus laws the applicability of which depends on recognizing when other things are equal. But it is notoriously difficult to specify such qualifiers explicitly. In fact, it looks like ‘other things being equal’ is kind of like jazz - you know it when you see it. I’m not sure if this is the same kind of worrisome mind-relativity that afflicts non-natural kinds though.

  10. Nick says:

    Pete, those are very interesting remarks and I look forward to your next blogging on this topic, whenever you get the chance.

  11. Pete Mandik says:


    I think you raise some pretty interesting points.

    Regarding the how many nomic relations question, in theory, one would expect that there is a range from things that enter into one nomic relation, things that enter int two, and so on. Why not? However, this isn’t how things actually turn out as far as I can tell. I’m likely just pulling this out of my hat, but it looks to me like things pretty much just fall into two categories: things that enter into just one and things that enter into a whole bunch. And by a whole bunch I mean something vague like, a lot more than two. Further, the things that enter into just one enter it into nomic relation only with us. If this is indeed the truth of the matter, it cries out for some kind of explanation, and perhaps one can be given.

    things that enter into none won’t be noticed by us because they can’t be noticed by us. They would be, by definition, undetectable. Things that enter into just two where one is with us and the second is with something else are unlikely to get on our radar because there’s unlikely much of an interesting science to be had. Same for 3, 4, and up until you get to n. But I don’t know what n is. I guess you get a bit of a sorites, here. There’s room, as you suggest, for a range of naturalness. But I still think there’s such a distinction between the natural and the nonnatural and its a bit of a sorites puzzle to say how much is enough.

    Regarding the ceteris paribus point. Perhaps this is a basis for an argument for reductive physicalism? If ‘other things being equal’ is pretty much like jazz, then the ceterus paribus laws are more about us carving the world than the world providing joints. If the laws themselves are natural, then they better involve kinds reducible to (complex combinations of) physical kinds.

    The above is all super-ultra-highly-speculative and if I won’t be surprised if it turns out to be totally wrong.

  12. Tad says:

    Pete -

    I think you’re right that it could be a basis for reductive physicalism. Or it could throw doubt on the whole opposition b/w natural and non-natural. Maybe some kinds the identity conditions of which make reference to our abilities to identify them still generate real patterns interesting enough to study scientifically. This gets back to my earlier skepticism about your claim that non-natural kinds aren’t suitable subjects for science. There’s a kind of dualism implicit here (not metaphysical mind/body dualism, but more like natural/artificial dualism) that I think might be on the wrong track.

    According ot this view, anything sullied by requiring our responsive definitions to be identifiable isn’t something natural or worthy of study by science. But surely we’re natural and so are our responsive dispositions. Why can’t kinds that require our responses to have some kind of integrity and efficacy also be natural, and studiable by science? Consider the effects of fossil fuel technology on the environment (mentioned a few posts back).

    I still think Dennett’s more inclusive approach makes better sense of actual science. Facts about the world, independent of our responses, make it the case that systems that can only be defined in terms of our responses give rise to robust patterns that science can study. Teleological, intentional, computational patterns are real, and robust enough for science to study, b/c natural facts make it the case that the actual and counterfactual behavior of certain systems is robustly trackable when treated as purposive, intentional or computational. All of this can supervene on natural facts, even if part of the supervenience base includes our responsive dispositions.