I once gave a (perfectly awful) cognitive science lecture at a major centre for brain imaging research. The main project there, as best I could tell, was to provide subjects with some or other experimental tasks to do and take pictures of their brains while they did them. The lecture was followed by the usual mildly boozy dinner, over which professional inhibitions relaxed a bit. I kept asking, as politely as I could manage, how the neuroscientists decided which experimental tasks it would be interesting to make brain maps for. I kept getting the impression that they didn’t much care. Their idea was apparently that experimental data are, ipso facto, a good thing; and that experimental data about when and where the brain lights up are, ipso facto, a better thing than most. I guess I must have been unsubtle in pressing my question because, at a pause in the conversation, one of my hosts rounded on me. ‘You think we’re wasting our time, don’t you?’ he asked. I admit, I didn’t know quite what to say. I’ve been wondering about it ever since.
See also:
Grill-Spector, K., & Weiner, K. S. (2014). The functional architecture of the ventral temporal cortex and its role in categorization. Nature Reviews Neuroscience, 15(8), 536-548.
ABSTRACT: Visual categorization is thought to occur in the human ventral temporal cortex (VTC), but how this categorization is achieved is still largely unknown. In this Review, we consider the computations and representations that are necessary for categorization and examine how the microanatomical and macroanatomical layout of the VTC might optimize them to achieve rapid and flexible visual categorization. We propose that efficient categorization is achieved by organizing representations in a nested spatial hierarchy in the VTC. This spatial hierarchy serves as a neural infrastructure for the representational hierarchy of visual information in the VTC and thereby enables flexible access to category information at several levels of abstraction.
“It belongs to understanding how the engine in your auto works that the functioning of its carburettor is to aerate the petrol; that’s part of the story about how the engine’s parts contribute to its running right. But why (unless you’re thinking of having it taken out) does it matter where in the engine the carburettor is? What part of how your engine works have you failed to understand if you don’t know that?”
ReplyDeleteFodor overlooks a key aspect of researching consciousness when he makes this analogy. Here, it is easily stated that the location of a piece is irrelevant to understanding the function of the whole. But that is not so easily stated. First, a key question about consciousness is precisely what it is. We don’t understand how or why it allows us to do what we can do. Knowing that something happens or can happen (or knowing that we do have mental states) does not translate to knowing how and why it does, which is the easy question of cognition. Fodor also assumes that building a “brain” is not the key to reverse-engineering consciousness; understanding the location of the parts will not help understand the whole. I would argue that by performing these studies, it can be just as important to identify what doesn’t contribute to understanding and higher consciousness as much as knowing what does. But this would also require a clear and rigid definition of consciousness in order to separate it from other mental tasks, a definition which has not been set and, it could be argued, would be impossible to set. This same issue contributes to the difficulty of studying action “understanding” in mirror neurons, since there is no definition of what action understanding even means.
Fodor is saying that finding out where and when brain activity occurs will not tell you how and why it produces the capacity to do what we can do. That's the easy problem. Consciousness (the hard problem: how and why does the brain produce feeling) is another matter, but the answer is the same: where and when won't tell you how or why.
Delete“As far as I can see, it’s reasonable to hold that brain studies are methodologically privileged with respect to other ways of finding out about the mind only if you are likewise prepared to hold that facts about the brain are metaphysically privileged with respect to facts about the mind; and you can hold that only if you think the brain and the mind are essentially different kinds of thing”
ReplyDeleteReading this as a cognitive neuroscience major was especially painful. I don’t think it’s fair to call every neuroscientist a dualist simply for caring about the brain. Localization of function is extremely important when it comes to the study of disorders. Understanding normal functioning is a key part of understanding dysfunction. It’s important to know that damage somewhere near the back of the brain could impair vision both to know where to look in patients with vision problems, and to know where not to poke during surgeries.
Now that I’ve pitched my mini argument from a neurological standpoint, let me try to address the philosophical aspect. I’d argue that all of these different cognitive aspects (decision making, language, …) contribute to consciousness. Maybe we only “feel” when most of these cognitive aspects are present, and this is why we start wondering at what point people in vegetative states stop being considered conscious. Something interesting that could come from mapping is that by knowing which areas (and hence which functions) are connected, it gives us insight into the order of how we could be processing things, hinting at ways that we could reverse engineer this into AI. In that sense, it’s important to know how we do things so that we can figure out how to get a robot to do the same things. Is it possible to get a robot to do the same things through different mapping? Maybe, but if we want to understand specifically how we cognize, then it’d benefit us to try to replicate what we do as close as we can.
Fodor is talking about cogsci and the easy and hard problem, not about clinical neuroscience. No one doubts that studying the brain is necessary if there's something wrong with it and you hope to fix it.
DeleteBut you are right that Fodor, a philosopher, like Searle, a philosopher, likes to accuse people of "dualism" for no particular (or relevant) reason. His critique of neural imaging (in cogsci) is not about consciousness (the hard problem) but about whether it helps solve the easy problem.
Fodor's message is simple: When and where won't tell you how and why.
Could you explain or give an example of what you have in mind with "knowing which areas (and hence which functions) are connected, it gives us insight into the order of how we could be processing things"? (And what do you mean by functions? (Or do you just mean areas whose activity is correlated with sensing, moving, remembering, feeling, etc.? Fodor's question there would be how those correlations can help us explain causally how to produce the doing-capacity that comes from exercising those "functions.")
It is clear that Fodor understands the importance of knowing where specific cognitive functions are generally located in the brain from a medical standpoint, as noticeable in his language and surgery example. It seems what he questions is why knowing the exact location of a function in the brain would reveal information about how the mind works. I agree with you Lyla when you say that location matters when your goal is to reverse-engineer cognition, especially since the path we seem to be taking is to recreate the human brain. I wonder, though, if Fodor believes that we are becoming too specific and seemingly lost in detail when it comes to neuroscience. I strongly believe that neuroscience is important for understanding cognition, and I bet Fodor does too, but I think he has a point when he suggests that a lot of neuro-imaging studies do not connect to how cognition, or the mind, works and are instead distracted by finding the “exact” local of a specific function which may not be particularly relevant.
Delete@prof I'm talking about cognitive functions so not quite sensing and moving but more so remembering and decision making. I feel like it's easier to create a robot that can sense using light sensors and move using little motors (is that how robots work?) and in that sense memory and decision making can be pretty easy for a robot just by using probabilities. But we know (through brain mapping) that other things come into play for us when making decisions, like reward sensitivity and emotions for example. Knowing this helps us factor in other elements to make the robot's decision making more "realistic". I think it's less about knowing where certain regions lie and more about how they are connected to one another. In terms of when and where not being able to tell us how and why I'd say we're getting close to the "how" since maybe the connection of all of these areas is what makes it feel like something to remember for example. I'm sure some would argue that we're answering the "why" by saying these exact combinations of activity in these circuits are what make remembering feel like something, but that doesn't sit too well with me. Saying why we feel is due to neuronal activity sounds like something someone who insists on T4 would say. I know it feels like something to remember that robert's favorite third grade teacher is madam marquise (although I don't know how to spell her name), I don't know why, but maybe how is through those neuronal connections. Can I say with certainty that these circuits are how we come to feel things? No but we can try. There are other approaches we can take before deciding to study the brain (although Searle believed otherwise), but it's there and neural imaging is helping us understand it, so might as well use the knowledge we're getting from it to try and solve the easy problem
Delete@matt my fragile almost-neuroscience-degree-holding soul would like to believe that fodor thinks neuroscience is important for understanding cognition but, to me, it sounds like he thinks we're just wasting time and money. In our defense though, I don't think neuroscience has been about finding exact locations of functions for many years now. We've moved past the idea that brain region x is responsible for function y a while ago it feels like. Now it's more about connections between region x and w and z contribute to aspect 1 of function y by modulating some signal or another. Saying the ventromedial prefrontal cortex is what is responsible for assigning values to rewards is a flawed statement. Imaging cares about the inputs the vmPFC is getting and to which regions its projecting and how it's affecting those other regions. All the neuroanatomy courses I've taken have been about circuit mapping as opposed to identifying regions and I think that speaks volumes about how the focus of the field has shifted, showing that neuroscientists also believe there's more to the story than simply region x causes cognitive function y. Whether all of these studies are looking for an answer to the easy problem is another case, but perhaps the answer to how cognition feels like something lies in the complex interactions of these neural networks (and perhaps not, in which case we'll just have to settle for the clinical contributions)
DeleteMatt, if we keep the "easy problem" of doing-capacity clearly in focus (and don't mix it up with the much harder "hard problem" of feeling-capacity) then it will be the contribution of when and where to successfully explaining how and why that will determine whether Fodor's skepticism was right.
DeleteLyla, I think you may be prematurely wrapping bits of the hard problem into the easy problem (making them both the hard problem).
But remember that the "easy" doing-capacities of the T3 cannot be accomplished by a computer on wheels with an optical transducer.
And "why" is just as causal and objective a reverse-engineering question as "how": It is asking about the adaptive value of the capacities that organisms have: why can some fly and some swim and some speak? It's not particularly a question about the hard problem (how and why can they feel?)
So far (as far as I know) no cognitive or behavioral capacity has been successfully reverse-engineered as being caused by "connections" -- except in the very special case of neural nets ("connectionism"). Trouble is that almost all neural nets we read about are actually computational simulations of neural nets -- so they don't really even have those connections!
As for "circuits": we know how electrical circuits can do what they can do; but so far saying that some capacity is caused by a "circuit" leaves us clueless as to how the circuit does it as it does to say it's done by an area (except in the case of sensorimotor reflex circuits).
I must say, I particularly enjoy reading Fodor for nothing more than his verbose sass. A favorite for me was in his response to Bly when talking about the difference between finding the roots of the cognition and looking at the mind: "Roughly, it’s the difference between a scientist who has a hypothesis and one who only has a camera."
ReplyDeleteIt appears to me that Fodor is deeply attached to the scientific method - and I'm not sure why. Put concisely, my point is that we do not have to gather data with a question in mind, even if our overarching goal is to answer questions. Why? Not all questions are asked or answered in the present: many questions are asked only when more information has come to light, and answered using data collected in the past. It seems to me that we are populating database filled with all the mind-brain connections.
So, it seems to me quite alright to answer to Fodor's inquiry, why do you take these pictures, with the same reasons that detectives have for taking pictures at a crime scene: "It's to provide the answers to questions we don't yet have".
I certainly second your enjoyment of Fodor's writing style.
DeleteAnd I would follow up your argument to the benefit of obtaining the information with the comparison to the noted benefit of looking for information on cognition in a variety of places - computation, brain imaging etc.
We do not know, and therefor gaining these further insights that may spark ideas or that may provide insight when we produce them elsewhere is not a bad thing.
Although I will grant him the point about the costliness of brain imaging and the funding of projects. I do see the argument that perhaps there could be areas where the money is better spent. However Fodor never actually provides us with any alternatives to fund.
I really like your analogy of taking photos at a crime scene and I think it is a productive metaphor in more ways than one. Certainly we can take photos for future reference, but also photos are not the only things collected from a crime scene...lots of other measures and samples are taken too! I don’t think that we will be able to learn everything about the how and why of cognition just from neural imaging showing localization, but I think it is a useful piece of evidence alongside other observations from psychological experiments, linguistic research etc..
DeleteOne thing I think Fodor misses is that neural imaging doesn’t just tell us the where, but also the “how much.” For example, in the literature on decision making, there is an idea that to make decisions we look at the subjective value of various options, and these values (or the differences between them) can be correlated to the amount (the "how much") of metabolic activity in the brain (see source below). Can I just look at the neural imaging and say I know how we make decisions? No, but it can be one piece of supportive evidence for a hypothesis on how, in this case the neuroeconomic idea of common currency.
Levy, D. J., & Glimcher, P. W. (2012). The root of all value: A neural common currency for choice. Current Opinion in Neurobiology, 22(6), 1027–1038. https://doi.org/10.1016/j.conb.2012.06.001
Adding to this, scientific research often begins with data gathering. We have a general idea of the question we are investigating, but very often trends emerge once we already have the data and guide our line of thought. We observe correlations and draw conclusions. If our aim is to understand how cognition works, the more we observe the brain, the more we can make sense of its different components and its structural specificities (if any) and how they are conducive to computation (or whether they are irrelevant). At some point, the dots may connect and we may come up with a revolutionary theory. Alternatively, we may eventually discard neuroimaging as we did introspection. But just as the CRA led us to question computationalism by pointing out a flaw, if neuroimaging turns out not to give us the answers we need, it will at least force us to look in another direction.
DeleteJG, you may be right that just gathering data like a detective may eventually lead to a hypothesis, but trying to get the answer to "how and why" from "when and where" does seem like a tall order. And although there isn't really a "scientific method," in practice it's usually gathering evidence to test a hunch. (Didn't Sherlock Holmes have hunches, rather like Galileo?)
DeleteDeirdre, you're right Fodor provides no alternatives: Searle offers studying the brain as the alternative to computationalism, but Fodor (who was nominally a computationalist, but without actually doing any computational modelling) seems to be saying there's nothing to study if you turn to the brain: You need brain data if you want to explain T4 capacity, but T4 has to include T3 capacity, and you can't explain that from studying where and when things happen in the brain: That works for the heart, but not the brain, because unlike the heart, what the brain does (inside) is too different from what it does outside: what the brain does outside is everything we do!
Stephanie, yes neural imagery can tell you about (correlations with) location in time and space, but also intensity. I suppose neural activity intensity could be used as another psychophysiological measure of behavior, like a lie detector. That's good for predicting behavior, and even feeling, which is helpful for mind-reading. But do you have a concrete example in mind of how it could help in reverse-engineering cognitive capacity? (Did mirror-neuron activity help us figure out how the brain can imitate movement? Didn't we already know it could, because we can! But we still don't know how. Nor would correlations with neural activity explain how we do Shepard's mental rotation any more than reaction time does.)
[The one who cannot be accused of not trying to explain how and why something can be done is the one who tries to build (or simulate) a device that can do it.]
Ishika, you are right that science (and reverse-engineering) are based on data (i.e., evidence). What is not evident is whether correlations between brain activity and bodily capacity will give causal explanation of bodily capacity.
Although I agree with a previous commenter that reading the Fodor paper as a (potential) future neuroscientist was a bit painful, I do think it is good to be faced with, and have to respond to, skepticism. What I do take issue with is that I don't find Fodor's points against the usefulness functional localisation to be very compelling.
ReplyDeleteFirstly, I'd like to point out that Fodor's impression of how functional activity is interpreted is a bit oversimplified. To say that the point of neuroimaging is to find certain regions of the brain that are responsible for very specific thoughts or behaviours like 'teapots' or the infamous 'Grandmother cell hypothesis' is missing the point I think. The importance of these correlations is discerning whether activation of areas for those very specific stimuli are parts of larger functional specificities and systems.
One overarching point or question that Fodor asks is whether functional neuroimaging is actually helping us unravel the mysteries of mental states but my question is can we really know how the mind works if we don't have any insights into the main machine of consciousness (the brain)? Not every technique in neuroscience is suited to tackle these questions, but couldn’t functional neuroimaging help us understand how to better reverse engineer human capacities? These questions about the human mind aren’t only theoretical or philosophical and I think it’s crucial that we know the insights into the physical structure that these technologies can provide.
Building on this, Fodor says "In particular, brain imaging is expensive compared to other ways of trying to find out about the mind." I am curious as to what other ways he is referring to (which have not been discredited in history as not yielding the answers we wanted about the mind). I agree with Allie that as of right now they are our only insight into the neural processes going on when we cognize/are conscious, and discarding this method means potentially discarding important information that could help us reverse-engineer cognitive processes - arguably the proof of having understood cognition.
DeleteAlso, relating this to our other reading on mirror neurons, fMRI studies were mentioned throughout the paper in several studies. Whether understanding mirror neurons is important for reverse-engineering or not, insights like these are central to understanding what computations our brain performs, or at the very least, can contribute to questions such as the one about hardware-independence.
Allie, you ask "couldn’t functional neuroimaging help us understand how to better reverse engineer human capacities?". Fair enough do you have an actual (or even a hypothetical) example of how it could (for a cognitive capacity, not a vegetative one, like breathing or temperature regulation).
DeleteIshika, you are right that Fodor does not give alternatives (although he was a computationalist). But how does imaging, and finding mirror-neuron correlations with imitation capacity help us "understand what computations our brain performs"? And how can the brain teach us anything about hardware-independence (which we know is true of computers but not of the brain, unless it's just a computer, computing)?
I'll try at a hypothetical example of how functional neuroimaging could help us understand how to better reverse engineer human capacities. This also relates to William's post a little lower.
DeleteFunctional neuroimaging allows us, as Fodor pointed out, to see the centres of activity related to an action or a thought. As William pointed out, it also allows us to see potential connections between brain areas. Thinking about teapots might make us see an activation in the olfactory system of the brain (thinking of teapots makes you smell the tea they contain), visual cortex (thinking of teapots makes you imagine your grandmother's teapot), basal ganglia (thinking of teapots makes you think about your tea-drinking habit), and so on.
Fodor seems to purport that knowing where brain activity occurs in relation to certain events (i.e. thinking, doing) is useless to us. But do we not aim to reverse-engineer a person? Can we deny that there is cause and effect between our "engine" starting and us going forward (e.g., completing an action)? If what a person is doesn't include what they're made of, then Fodor is advancing computationalism, and Searle has showed us (at least in some way) why that can't be the whole answer.
We're trying to find out what makes a person, and in my humble opinion, any avenue that might give us some information about what makes a person tick can't be too far away from the goal.
Eli, see Replies above about connections, circuits and reverse-engineering causal mechanisms.
Delete"Granted that we always sort of knew that there’s a difference between nouns and verbs, or between thinking about teapots and taking a nap, we didn’t really know it till somebody found them at different places in the brain. Now that somebody has, we know it scientifically."
ReplyDeleteFodor's questioning why we are so interested in learning what happens north of the neck was striking. We are attempting to reverse-engineer the brain so it can pass a Turing test, where a computer would be indiscernible from a human being. Is this because we want to know how to recreate cognition in systems other than humans, or because we want to know just how far we can go with our robots before creating creatures able to cognize on their own? Are we just trying to prove that consciousness exists in others than ourselves (going back to the Other Minds problem)?
Cogsci is trying to solve the easy problem. Fodor is doubting that brain imaging correlations will help. (And that'd before even mentioning the hard problem!
DeleteI wanted to take a stab at a point brought on by others in that the location of certain cognitive functions is the brain can inform us of how the mind works.
ReplyDeleteI absolutely concur with Fodor in that taking note of the areas where the brain lights up during a particular activity is only useful in a medical sense as previously mentioned and serves little purpose for determining how the mind works. The accuracy of this data is not completely accurate anyway because brain imaging implies the existence of the general mind where everyone has the same functions associated with the same areas. In truth, there is individual variability, e.g. some people have language lateralised to the right side of their brain rather than the left (which is where it’s noted as being concentrated in the literature).
Where this information is not useless is considering where areas light up in relation to each other. This is because areas that light up together could imply that a particular connection between the areas is being used. Connections in the brain are not homogeneous and are each supported by different neural architecture. If the way that neurons are connected differ according to each circuit, doesn’t this mean that the algorithms that are realised by these same circuits can differ? Furthermore, if we gained sufficient knowledge about these particular circuits, wouldn’t it be possible to extract the algorithms that the circuit affords such that T4 wouldn’t be necessary and we could transfer them to T3?
Hey William! I want to take a stab at your stab haha
DeleteI really like what you bring up about the potential connections between areas. So far throughout this class, we've talked a lot about how it's not about the individual differences between brains matters, but it's about the very very fundamental bare bones of cognitive mechanisms. Maybe brain imaging helps us by showing that there are "universal" (I use the term lightly) algorithms/circuits that are common among everyone. And this could give us a hint towards the easy problem?
William: are you still confident that neurons are executing algorithms? And do you think you could figure out what algorithm a computer is executing from the timing and location of its activity?
DeleteEsther: If what and where were uniform across all people, would that help with how and why? (How?)
I think finding what and where the patterns occur could help with the how, because different areas of the brain have different neurons with different properties? so this would give us an idea of how the circuits are related to one another... but i'm not too sure about the why.
DeleteDoes our "why" answer need to stand alone from the how...? Like when the authors for the associative account for MNs share that they want to separate origin and function. is that what we want to do with thinking as well?
Professor:
DeleteTo answer your second question, I believe that knowing the location and the timing of an activity of a particular segment in a computer doesn’t afford us much knowledge, but knowing the location and timing of a computer segment in relation to others allows us to infer information about the implementation of the algorithm e.g. which circuits connect which other, the weight of each connection, etc. While an algorithm can be multiply realisable, the way a circuit is connected constrains the types of algorithms thar are possible to be implemented. This relates to the point that Esther makes about how differential properties of neurons in each circuit could constrain the possible algorithms that the neurons are implementing. But I’m not certain if this would allow us to identify the specific algorithm at play, no.
Esther: Explaining "how" would be enough; but see Replies on connections and circuits. On "why" see other Replies: evolutionary explanation, adaptive value.
DeleteWilliam, do you think neuroscience is about finding out how the brain implements algorithms?
Hi all!
DeleteI definitely believe in the potential informational value (reduction of uncertainty) in identifying significant connections between different electrical and other modal signals, occurring in different cortical locations, during different activities/feelings we can “do.” I would argue that mapping out multi-modal correlations in time and space can help produce a greater understanding of the "how" and the "why" in the easy problem.
Can we take successive images or video footage of certain neural tissues/areas and slow it down such that we can observe the chronological paths of electric currents / action potentials, from neuron to neuron and area to area? Can we observe the path of signals, for example, from the moment our photoreceptors perceive light stimulation, follow the polarity changes that affect the flow of electricity, and then watch the waves of activity travel successively through the various cortical areas classically associated with generating vision? Would that give us a satisfactory explanation of "how" the brain and eyes can "do" visual-recognition? Even if the imaging can illustrate and thus help us map the circuits and pathways of signals in particular brains, it does not seem to tell us much about how it is that it “feels like something” to see.
Above in earlier comments, Lyla suggested that "maybe the connection of all of these areas is what makes it feel like something to remember for example" to support the idea that knowing the "where and when" of neural activity through neuroimaging techniques can help us understand "how" it is that neural activities generating memory or visual-recognition feel like something when we "do" them. Prof. Harnad responded that she might be mixing the hard and the easy problems - but isn't the easy problem "how we can do what we do," and feeling is a thing that we can "do?" Then, although it may sound like arbitrarily jumping to the hard problem from the easy problem before solving the latter, doesn't it still help with our goal of trying to explain how and why we all feel like /something/ while our neural nets are doing what they can do? We can connect correlative "feelings" with certain whens and wheres of different neural activities, and then perhaps further down the road, with much larger samples of data on such correlational trends, we could start to recognize or identify more specific causal mechanisms linking x-activity to y-feeling-of.
Compared to the reading for 4a, Fodor’s piece in this journal was so much fun to read that I got to the end and couldn’t remember what the connection was supposed to be with our class.
ReplyDeleteOne thing that I’d like to address first, although it doesn’t have anything to do with MNs or cognition, is that I agree (although maybe not as harshly) with Fodor’s despair that money goes where the attention and spotlight is. I feel like there is a tendency now in the sciences to only ever share the “success” stories, without acknowledging the failed attempts that come before them. There’s also this idea that “science truth is the ultimate objective truth” but I have read enough papers to know that this isn’t true – seriously, every paper has a section on “all the reasons why this might not have been the actual case”. So I can see why he might be frustrated with the “dramatization” of scientific papers.
In terms of the usefulness of brain imaging, I personally just don’t think that localisation of different functions in the brain will ever help us get close to solving the hard problem. In terms of the easy problem though: I don’t have anything spectacular to add, other than ask where else could we look? We know that introspection isn’t enough. So it’s gonna have to be inside the brain!
Computer scientists have figured out how to do a lot of things that formerly only brains could do, And they did it without looking at the brain. (Of course those things are just toys, not TTs, so they may have nothing to do with T4, or even T3. But they do show there are other places to look.)
DeleteI wonder whether equating toy computer capabilities with human cognitive capabilities depends on the degree of equivalency that is being required. For instance, if you define the phenomenon of speech by all of its characteristics rather than just the most salient ones, then you can’t truly say that “Siri” speaks. Of course, when we think of what defines speech we do not generally think of such things as neuronal activity in Broca’s area, and so it could be argued that this is not part of the definition of speech. However, given that the concept of “speech” exists only to describe the phenomenon as it exists in humans, and that toy versions of speech exist only in imitation of this phenomenon, one could also argue that it inherently encompasses all aspects of it. I guess that this comes back to the debate surrounding whether T4 should be required due to the fact that internal processes are also in a sense observable behaviour.
DeleteIn general, I feel that any explanation of cognition should be neurologically plausible. Otherwise, I feel that we are abstracting “cognition” from the original raison d’etre of cognitive science, which is to learn something about how humans think.
You would not consider yourself to have understood much about birds by endeavoring to reverse engineer flight and producing a plane. Obviously, the resemblance between a T3 robot and a human is infinitely greater than between a bird and a plane, but my point is that when cognition becomes defined as a type of behaviour rather than a neurological process, there is a risk that we could learn about the behaviours without learning about ourselves.
A social commentary to support Fodor's skepticism: more than ever before, casual readers have access to quality scientific research that they can read at the own leisure. More people now have higher education which allows them to digest this new form of entertainment that is popular science. I understand where Fodor is coming from, since he is a philosopher after all, not a scientist.
ReplyDeleteI agree with Esther about the "where else" issue. It's enticing to have images that give us some sort of concrete information, not just merely speculating. However I think brain imaging is the best tool we have at the moment. Though it surely does not answer the "why" and "how" questions, I can't think of any other way to address it.
Hi, this is a reply to both Esther and Wendy's posts. :) You both brought up some great points! As students, we rarely ever read about failed or inconclusive scientific experiments, even when they still provide us with valid results and valuable information.
DeleteThere are some cons to brain localization research. Fodor mentions that neuroimaging uses scarce resources, such as time, money, and computer power. As well, it is sometimes at high risk of false positives. (You can take the Atlantic Salmon fMRI study as an example!)
In terms of cognitive science, I agree that brain imaging will not be able to solve how/why can we do what we do, or how/why can we feel the things we feel. Nonetheless, I am optimistic about its future implications. It is a relatively new technique, and like many other technological advancements, it will continue improving, updating, and providing qualitative and quantitative data about the brain for other scientific purposes.
Wendy: about "where else, see above.
DeleteTing, false positives for what? We are talking about cogsci research, not clinical research.
This paper sort of reminds me of Searle’s paper, where we criticized him for not telling us where to look to find consciousness, he only that we shouldn’t look at computation. Fodor seems to be doing the same thing, where he is simply telling us that looking at the components of the brain won’t get us closer to solving the hard problem. What does he suggest we do instead? Does he want us to just give up? The hard problem is not supposed to be easily solvable, that’s why we call it “hard”. Neuroscience is a very young field and we have hardly given it a chance. I think studying the brain is the best chance we currently have.
ReplyDelete(Also, I am hoping we can hear from someone who is not a white man soon. What is up with all these white male cognitive scientists who are so full of themselves? Cognitive science is a new field, I hoped it would be a little more diverse!)
For the easy problem we are 'just' looking for cognitive capacity, not consciousness (feeling).
Delete(No gender has made much progress yet on the easy problem, let alone the hard problem...)
Contribution from people with diverse experiences can only be beneficial to the field. Similarly, learning about the field through the works of a variety of individuals can only benefit us as students. I, too, hope we'll hear from someone who's not a white man soon.
DeleteI wasn't insinuating that women have solved or could solve the easy problem, I was saying that it can be worth hearing from people from different backgrounds (ie not only white men). Regardless of the progress that has been made in this field, why have I heard so little from women and people of colour? This is not a criticism of this course, it is something I have seen repeatedly in cognitive science.
DeleteIt seems to me that Fodor is criticizing neuroscience for just gathering correlational data on brain activity and behaviour via imaging techniques without a semblance of a paradigm. In other words, neuroscience is just piling up images that tell us where function is localized but not how those functions occur (at least not in terms useful for reverse-engineering). He doesn't deny that this is useful for things like brain damage but the excitement doesn't match the output. We won't explain cognition by just doing that. Without some sort of guiding hypothesis that might allow cognitive science to formulate principles of cognition, then it does seem that neuroimaging might not be so helpful. But Fodor doesn't really give an alternative. Why the brain? Maybe because there is nothing else that does what the brain does?
ReplyDeleteThe "how" is not about "how those functions occur" but about how those functions produce the capacity they're supposed to explain.
DeleteI also questioned extending Fodor’s argument to the brain in general – as the organ where cognize/feel states are localized. Fodor places emphasis on whether, rather than where, all that we can do can be localized in the brain, and he speculates that perhaps so much investment in seemingly irrelevant neural imaging research is fueled by the desire for conformational assurance. In other words, many seek to scientifically support and confirm phenomena that seem rather intuitive. I’m curious to know if this was always the case with the brain. Didn’t we gravitate toward the brain following anatomical discoveries? Or was studying the brain always understood to be the center of cognition, and we simply sought scientific confirmation?
Delete“Modular minds are at best hypothetical and somewhat improbable, but evidence for the modularity of aspects of brain processes - vision is a good example - is strong.”- Steven Open in response to Fodor
DeleteI found Open attempts to nuance Fodor's argument to be interesting. Open, in agreement with Fodor, argues that while parts of the brain may light up when we do something like thinking about teapots, we can't claim that thinking about teapots is localized to that particular lit region. However, he adds, in opposition to Fodor, that rather "all we have done is found part of a system enabling us" to think about teapots. Does Open's response really add anything to Fodor's argument? And, is "enabling" our ability to experience felt states not part of a causal mechanism? I agree that we can say that visualized brain activity isn’t just cognition; but is it even necessary to pass the Turing Test/necessary for T3?
“With occasional anomalies, the argument between homogeneous minds and heterogeneous minds aligns with the argument between empiricists and rationalists”
ReplyDeleteI know that this debate has come up already and that it wasn’t specifically what Fodor was going for with this quote, but debating some epistemics in cracking the easy problem of cognition seemed particularly relevant to this problem.
Hume established that proving through induction, which pretty much resumes the traditional scientific method, cannot give us a universal law-like property (as opposed to deduction), because induction works on predictions conforming to past regularities. In other words, this sceptical position defines scientific methods as being merely correlational, because not all generalisations can be confirmed by their instances. It feels to me that what Fodor is raising inscribes itself in this very long tradition of philosopher doubting the scientist’s methods of induction in that finding various instances of brain states by neuro-imaging will not give us a sufficiently universal theory of processes of our minds. I genuinely do not know where to position myself on this matter but I recently read a provocative article (for which I could no longer find references) that quite cynically predicted the obsolescence of causation to correlation and gave a hint at the benefits of neuro-imaging. With the advent of machine learning swallowing masses of data scientist are now mapping entire genomes using machine learning rather than hypothizing and testing one by one the expression of genes. What I’m saying is that instead of having a genius come up with E=MC^2 all you need to do is simulate enough experiments with enough data and precise enough algorithms to get a correlational relation between matter and energy without ever having to go through a general formula. While these methods of correlation can have serious implications with false positives and we should not give up on our capacity to prove causal relations, correlational techniques are essential on shining a spotlight on promising hypothesis. Reminding me of this article simply made me think that finding a causal explanation to how brain states cause mental states using correlational techniques could be a promising way to move towards a finer hypothesis. Whether or not the cost of finding this hypothesis through neuro-imaging is too high, that is beyond me however.
Descartes, too, noticed that only mathematics (formal deduction) could give you truths that are provably necessarily true (on pain of [formal] contradiction.
DeleteBut you can reverse-engineer causal mechanisms by "induction." That's what all of science is based on, including cogsci.
But Fodor's injection of nativism/empiricism into the question of whether when and where can reveal how and why (other than for simple I/O reflexes) is as irrelevant as his injection of dualism
Gene mapping, as far as I understand it, is structural. And structural correlations can be informative in themselves. (Unsupervised learning.) But they don't tell you much about function.
Neural imaging is good for learning brain anatomy, and even some functional physiology, but not for reverse-engineering cognitive capacities.
Fodor's argument about the pointlessness of mapping brain function can be capture by his comparison, "But why (unless you’re thinking of having it taken out) does it matter where in the engine the carburettor is?"
ReplyDeleteBut I'm hesitant to agree with him here because I think that understanding the physical functionality of the brain could potentially give insight into the mind, or at minimum, how the brain and mind interface. I might not care about where the carburettor is in a car I'm driving, but if I need to reverse engineer a car in order to figure how what it does and why it does it (Cog-Sci), it will definitely be valuable information to know where the carburettor is because it hints at it's functionality. If it's physically connecting the engine and the fuel pump, I'm one step closer to figure out what the machine does.
Fodor says near the end of the article, "What if, as it turns out, nobody ever does find a brain region that’s specific to thinking about teapots or to taking a nap?" and this shows that he doesn't understand that the goal of brain imaging research is not to figure out which parts of the brain map to which activities. The goal, or at least how I understand it, is to figure out how the brain functions, and figuring out which regions are active during different behaviors is the first step.
If we think about how brain mapping technology will advance, I think the goal becomes more clear because we won't be limited to a macroscopic view.
I agree with you in that if we're trying to reverse engineer a car exactly as it is, it's important to know where everything is. But I think it's worth considering the fact that we're not trying to reverse-engineer the brain, we're tring to reverse engineer cognition. So the question is if there's a way to do what the car does exactly and indistinguishable from other cars making an identical carburettor, are we failing at reverse-engineering what a car does?
DeleteI do think there might be cases where brain data is useful in reverse-engineering an aspect of cognition but I think what Fodor's point was that blindly looking for data and trying to make sense of it is a waste of time in regards to cognition, especially since that's not data we need to rely on and getting that data is expensive and pulls focus away from other, possibly more useful, research for cognitive science
Hi Sam! You make an interesting point. I think Fodor would agree with you that understanding the function of the brain could give us insight into the mind or mind/brain interface, but I think he would respond by saying that the problem with imaging is that it doesn't give us insight into function. As you and Fodor point out, imaging tells us what areas of the brain are active when a person is engaging in a given task or thinking a given thought. However, simply knowing that something is happening in a given spot doesn't really tell us anything about the causal relations that hold between the neurons that are active or give us information about the algorithms the brain is performing (if indeed it is doing computations). Hence, imaging doesn't give us the kind of causal explanation we should be looking for.
DeleteI think you make a very fair point which is that knowing where things happen could potentially be the first step towards understanding function because knowing that a region responsible for one task is in close proximity to a region responsible for another could hint at the functions and causal relations which hold between these regions. I think, however, Fodor might say that while this may be a small first step towards understanding function, this is the only step that imaging will ever give us. Imaging can only ever tell us where groups of neurons are active and at what times, and maybe this will give us a hint about connections of regions, but this one kind of hint is all we will ever get; therefore, the excessive investment into imaging research and the popular belief that imaging explains a lot about the mind/brain is unjustified. That is, the explanatory power of imaging is way overhyped.
I do agree, however, that it is never good to say a priori that a given bit of data won't lead to unforeseen insights. Discoveries can be surprising, so who knows what links imaging might allow us to make, even if, (if we agree with Fodor), it is unlikely it will yield anything very promising.
Sam, see other replies about brain and "mind" interface. ("Mind" is a weasel-word.)
DeleteThe easy problem is to explain how and why the brain (or any causal mechanism) makes us able to do what we can do.
Is reverse-engineering the brain like reverse engineering a car? That's like one of the questions on the exam.
Is the easy problem of explaining how and why organisms can do what they can do about "mapping"? or about "which regions are active during different behaviors"?
Reverse-engineering cognition requires a causal mechanism as an explanation, not a correlation between what organisms do and the patterns of activation inside them -- unless the patterns of activation ("when" and "where") somehow reveal the causal mechanism.
Ada Alex: what are the data we need to rely on to reverse-engineer cognition?
It's not about the causal relations between the neurons that are active but about how the neural activity (or anything) causes the doing-capacity.
Neural activity patterns (like responses and reaction times) are data rather than explanations. To make them into a test of a causal explanation, you need to have a potential mechanism. And if so, don't you think the most direct reverse-engineering test is whether that potential causal mechanism is able to generate the capacity itself, not just its reaction time or its neural correlates?
I wholeheartedly agree with Fodor in his argument that knowing the location of where certain brain functions are located does not give any insight into the question of how the brain works, or more specifically, the question of mind-brain relation. I think Fodor’s article is an important criticism to scientists who believe its use goes beyond clinical importance. The crux of Fodor’s argument is that we cannot explain the mind-brain relation by simply mapping out the brain, and to imbue brain imaging with greater importance than it has might be a mistake. But I also believe that this is a very broad question, and I do not think that neuroscientists are “wasting their time” with brain imaging. At this point, studying the brain is the best lead we have got when it comes to answering the question of how and why we think. If we find that certain mental states correlate with certain brain activity for the majority of people, isn’t this some sort of clue into how our cognition works? We are looking for answers without knowing the question, but perhaps if we gather enough answers, the question will one day come to us.
ReplyDelete"Mind" and "mental" are weasel-words. They usually just mean states that it feels like something to be in. No one doubts that those felt states are brain states, and that the brain somehow causes them. But remember the easy and hard problem: We want to know how and why the brain causes our capacities to do what we can do (easy problem) and how and why the brain causes felt states (the hard problem). The question is: How can neural imaging (when and where) help with reverse-engineering how the brain causes either of these: doing-capacity and feeling-capacity. It does not help to put these problems in the old terms, the "relation" between the "mind" and the brain.
DeleteI agree with you Alessia! Localization may not give us insight into how our brain produces cognition, but I think it's an important element of research, worthy of time and energy. I don't think there's any harm in gathering a multitude of information from a variety of angles. The question of what cognition is, how to reverse engineer it is complex, to say the least. I don't think it makes sense to expect that research will give us promising results right away. We still have so much to learn and the more information we gather, the more questions we will develop. I think that at some point we will have enough information to slowly start making connections. Even if certain lines of research don't seem valuable right now, who knows what insight they will offer in the future when we have more information and more advanced technology. I think it makes sense to study the brain in relation to cognition. As was mentioned above — where else are we to look?
DeleteAbout Fodor, J. (1999) "Why, why, does everyone go on so about thebrain?"
ReplyDeleteI think Fodor rightly points out that “when” and “where” neural activity occurs does not give an explanation of “how” and “why” we do what we can do. “When” and “Where” are just data points and in themselves, they don’t consist in a causal explanation. Nevertheless, the obvious fact is that what underlies neural activity or patterns of neural activity has to be the mechanism which explains cognition. This is analogous to saying that the brain produces cognition (which we all agree on). The question is whether neural activation or patterns of activation in the brain can ever shed light on the mechanism.
I would still think that doing a lot of “mapping the brain activity to cognitive abilities” might allow us to infer the underlying causal mechanism or at least give us some constraints to guide our inferences or to guide our investigation of the mechanism underlying cognition. Am I wrong in believing this?
"I would still think that doing a lot of “mapping the brain activity to cognitive abilities” might allow us to infer the underlying causal mechanism or at least give us some constraints to guide our inferences or to guide our investigation of the mechanism underlying cognition. Am I wrong in believing this?"
ReplyDeleteDepends on the reasons and evidence for your belief.
My take on this reading is that Fodor does not doubt the clinical importance of neuroimaging and that this technique certainly has its own significance and merit in medical practise (e.g. awareness of localization of language to avoid during surgery). However, Fodor does not believe that all of the funding spent on functional localization and neuroimaging furthers our pursuit in explaining our cognitive capacity. Discovering correlations where the brain functions in tandem with all the things we can do does not answer how or why we do these specific things. It would seem beneficial for Fodor to draw a clearer distinction in this regard, between the clinical and cognitive importance of neuroimaging.
ReplyDeleteThis makes me reflect on the challenge with the field of cognitive neuroscience. The brain is an organ of the body, so likewise with the studies of any other internal organ, its physiology, biochemistry, anatomy, etc. are necessarily studied as part of furthering our understanding of biology and medicine. However, the brain is twofold in purpose insofar as it also generates our entire cognitive capacity which encompasses everything we can do. As we strive toward achieving T3, it seems evident that we are capable of determining novel facts about the brain in the form of correlations between brain activity (where/when) and everything we can do. However, to reiterate, correlations limit us in our goal of solving the easy problem.
Jerry Fodor correctly points out that correlations between brain regions that light up and behaviour cannot tell us how we cognize. No matter how much we study neuroimaging, or even directly study the brain tissue that is associated with said behaviour, we are still clueless on the emergent properties of the brain (cognizing and feeling).
ReplyDeleteWhile this limitation is very real present and often ignored by scientists, there is not much that can be done about it at the present moment. Adjusting our research questions could be a solution, but it is hard to say that it could guide us any further. As it is not just cognition that is perplexed with the easy and the hard problem, but so too are other schools of thought (E.G. Pain researchers). They study nociceptors, pain receptors, synaptic mechanisms and pharmacological interventions more closely than some neuroscientists and still cannot answer how electrical signals result in behaviour and feeling.
I think cognitive science has more benefits from studying the brain and knowing how it works (not necessarily down to the physics of neuronal movement, but a general overview), than without this knowledge. Because it grounds their knowledge into a specific base and doesn't allow for much deviation as to where one has to focus their attention on when thinking about the easy and hard problem.
As someone who has spent the last three years (mostly) studying neuroscience, learning imaging maps and the names of certain parts of the brain and the functions that we suspect reside there, Fodor's article is a bit painful to read. That said, I agree with Fodor in stating it is the question of "whether mentally functions are neurally localized in the brain", not where they are localized, that truly matters. Personally, I would modify his question slightly to investigate how this localization occurs - considering not only questions of innateness (where functions tend to arise) but also emphasizing questions of plasticity (what happens when given areas are damaged, and how do those functions, in certain cases, rewire themselves in different regions).
ReplyDeleteI would argue that while the precise locations of given functions in the neurotypical brain are not relevant to the questions pursued by cognitive science - the easy and hard problems - the processes by which brains 'rewire' are surely part of the easy problem and required for cognition e.g. learning, memory, and emotional associations. Understanding these processes - particularly as occur in relationship with analog components and sensory experience may benefit from neuro-imaging, tracing, and single cell recording.
While I understand that brain activity as recorded in this way only demonstrates correlation - not necessarily 1:1 causation, thorough experimentation may be able to bulk up and contextualize these correlations such that we can use them to reverse engineer and build artificial models, from which to test hypotheses about learning, memory, and association.
In an attempt to bring together some concepts across the course, I want to bring attention to Fodor’s remark on localizing language function. There is no doubt among the comments here that what Fodor emphasizes is the correlational data that neuroscience provides, missing the causation necessary for reverse engineering. This highlights why studying the brain is much harder than studying the heart. What I would like neuroscience to explore is how language areas affect our categorization ability. How do language areas in the brain affect our symbol grounding? What damage specifically interferes with these capacities? Although we cannot know how there areas cause certain outputs, we can certainly connect many different facets of cognitive science inquiry to the study of neuroscience. I also want to bring neural nets into the conversation. If we can create simulations of the brain which can provide more specific answers to the ‘how’ we do what we do, then neuroscience most definitely holds an important part to answering the easy problem.
ReplyDelete