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:
Harnad, S. (2019). Can neuroimaging reveal how the brain thinks? In Casting Light on the Dark Side of Brain Imaging (pp. 9-11). Academic Press.
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.
This may be a bit off track, but I feel like one reason why neuroscientists are so attached to the idea of brain imaging is tied to the love that science in general has toward objectivity. To know things "scientifically", we often have this idea that we need to know it from objective methods which cannot be biased by our human subjectivity. Brain imaging is one way to study "objectively" higher cognitive functions, which we previously had only been able to study with the help of questionnaires, self report, or psychological experiments, all of which are often consider not "hard sciences".
ReplyDeleteFor example, in another class, we talked about brain imaging studies that are used to measure pain (which areas light up when someone report pain, etc). The prof mentioned that these studies had elicited a lot of interest, because it allowed to study pain without the measure of subjectivity that is usually always there (since pretty much the only way to know how much pain someone is in is to ask him/her) and in this way it gave more credibility to pain research. Which seemed wild to me, because when someone is in pain, the problem that we want to solve is their subjective experience, no matter what the brain data looks like. Similarly to what Fodor said about mental state in his article: if someone's brain doesn't light up like it "should" when they say they are in pain, their pain is nonetheless real.
Yes, scientific research (and in fact any information-gathering to reduce uncertainty about how to get lunch) relies on objective, public, repeatable, verifiable observations and measurements (also called "data" or "evidence") rather than subjective hunches (as in climate change-denial, anti-vax theory, creationism...)
DeleteFodor is not questioning the clinical uses of brain imaging. He is questioning its usefulness in explaining cognition. The question to reflect on is:
Can "when" and "where" something happens in the brain while it is doing something reveal "how" and "why" the brain does it?
A particular example is: "Does the discovery of mirror neurons help explain the causal mechanism of the brain's mirror capacities, e.g., movement imitation, vocal mimicry, meaning-understanding or empathy?" We already knew the brain could do all that. Does finding out where in the brain correlated activity occurs help explain how the brain does it?" (T2, T3, T4)
Louise, I had a similar thought when reading this week's readings, this one in particular. Brain imaging gives a physicality to neuroscientists' scientific endeavours and research, which is certainly appealing, but I do feel like eliminating any and all subjective traces in explaining cognition will create a gap in our understanding. I would like to think there is a grey area between the extremes of completely objective measurements that provide data, and subjective feelings that give rise to conspiracy theories.
DeleteObjective measures allow us to “know scientifically” what we already know, which is great when we indeed already know something — for instance, that dogs salivate when expecting food. But I am having trouble wrapping my mind around how strictly objective measures can create and guide an explanation of how and why the brain does what it does. Mirror neurons are an objective finding; but they don’t explain the mechanism by which the brain imitates and understands.
Is it possible to go about trying to explain cognition using both objective tools and subjective markers? Or will this always be impossible due to the other minds problem?
I’m pretty sure that Fodor is suggesting that the easy problem alone is reason enough to deter a lot of research on the mind-brain relationship? Fodor is arguing that even if we find out where brain activity happens as a response to which stimulus, we will still not be able to understand how and why conscious beings do what they do. While I find this argument very compelling, I think that Fodor over simplifies the field of neuroscience a bit:
ReplyDelete“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.”
At least in the labs I’ve spent time at in my undergrad, there is a little bit more thought put into brain imaging than Fodor is suggesting here. Moreover, (as Fodor actually does mention in this reading) neuroscience and neuroimaging is not always primarily concerned with the question of how and why we do what we do. That being said, Fodor isn’t wrong when he writes that we are unlikely to understand the easy problem just through researching functional localisation by neural imaging. This reading definitely made me wonder why I have spent so much time over the past three years studying “neural localisation of mental functions” especially when I’m becoming increasingly confused about what a “mental function” is...
Fodor is critical of "cognitive" neuroscience, not clinical neuroscience (for which McGill is famous).
DeleteYou are right to ask yourself what makes a cerebral function mental (felt). But that's the "hard" problem, not the "easy" problem.
On Friday Fernanda will try to challenge Fodor's contention.
I think Fodor’s argument is an interesting perspective, though I feel he may take too strong a stance in the argument.
ReplyDelete“I want, to begin with, to distinguish between the question whether mental functions are neurally localised in the brain, and the question where they are neurally localised in the brain. Though I find it hard to care about the second, the first clearly connects with deep issues about how the mind works; ones that even us philosophers have heard of.”
I agree that knowing whether or not mental functions are neurally localised in the brain is important to figuring out how the mind works. But I disagree that knowing where is useless. Upon reading the article, I felt that knowing where might provide some kind of information that would benefit attempts at reverse engineering. Given that our brains have a specific layout, couldn’t it prove important for reverse engineering to take this into consideration in building a T3 passing robot? Upon watching the video linked, Professor Harnad clarified my overall sentiment when he noted that just because the knowledge of where hasn’t told us anything yet, doesn’t mean that it never will.
Can you give (or invent) a concrete example of how when/where could help reverse-engineer how/why?
DeleteHi Grace, I also had trouble with this, but I think it helps if you think of brain development from infants through adulthood. Brain regions are flexible and can change throughout development. Also, people can have brain areas in regions that others don't. For instance, people that are born blind usually have larger sensory areas for their other four sensations. These areas take over the region that's 'normally' the visual region in the occipital lobe. I think this would show that location doesn't matter because the functions can occur wherever they can.
DeleteNeuroplasticity -- the brain's ability to change, whether through development, learning, or response to brain damage -- is of course very important. But that too has to be explained -- and location is again not explanation.
Delete“If there’s a place in the brain where you find a whole lot of neurons going off when and only when whoever owns the brain is thinking about teapots, it’s at least plausible, all else being equal, that you’ve found where in that brain its thinking about teapots happens.”
ReplyDeleteThis quote really made me think of the concept of grandmother cells in the brain and how one cell could be responsible for whole objects, but only one category of objects. This can be contrasted with what we know about the brain already, many processes happen in multiple places which communicate with one another to comprise the whole.
On another note, I think the fascination with pin-pointing locations of specific function, specifically with functional imaging, is that it provides a visual simulation of what portions of the brain correspond to specific brain states or processes - which are not available visually any other way. Humans are visual creatures and I believe that being able to provide a picture or video of the brain allows us to more concretely contemplate the subject of the brain.
Yes, visualizing brain activity correlated with what we are doing or feeling is exhilarating, but is it explaining?
DeleteRemember that correlation is not causation.
I think although we can't reverse-engineer understanding, maybe we can reverse-engineer perception. In brain lesion studies or studies with TMS as in 4a, disruption of a specific brain region indeed disrupts the function of that region, and activation of brain regions helps restore their functions. I think this at least provides a causal relationship between perception and brain. Note that by perception I mean more than just sensorimotor input which can be acquired by camera and sensor of a T3 robot, but not to the level of understanding. And then we can build models that are physically similar brain structures responsible for perception. But maybe this is not that important for Cognitive Science.
DeleteTo your point about reverse engineering perception, I had just read this really interesting article about a neuroscientist who developed a gene therapy strategy to restore sight in patients with a specific retinal disorder (https://www.fastcompany.com/90674883/a-new-vision). As I was reading it, it made me think about this class and the never-ending tension between the what/where and the how/why. While cognition is possible without it, understanding vision is an important aspect in understanding certain cognitive functions. I do recommend this article, it was a great read and goes into some of the fascinating ethical questions in the medical/scientific community right now.
DeleteReverse engineering sensation is somewhat easier and much less cognitive than reverse engineering perception. Robotics and computer vision have tried "constraining" their models with T4 properties, but as far as I know this has weighed down robotics and computer vision rather than accelerating its progress.
DeleteI think the discussion of reverse-engineering ‘sensation’ or ‘perception is showing the question exactly what cognitive scientists will ask. Sensation is to the level of sensory detectors, it is much easier for a computer to be able to accurately ‘sense’ a stimuli. However, on the other hand, perception itself includes cognitive traits that trying to make sense of sensory information and provide information on how we can deal with it. Therefore, I doubt we can reverse-engineer perception. And I’m also not fully understand if there can be a state in between sensation and perception, as Zilong mentioned?
DeleteA quote that stood out to me was as follows: “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? Would that seriously be a reason to doubt that there are such mental states?” I think that this is an interesting question, in the sense that it would be rather impossible to actually find the brain area for every single mental state. However, by my understanding, finding the brain region for every mental state is not the goal of cognitive science. If we are interested in reverse engineering the brain in order to understand how and why humans cognize, then locating the brain region of certain mental states will not actually tell us much. However, creating a sort of map of cognition could be a good starting point in beginning to understand how cognition works.
ReplyDeleteHow does a map explain how? Can you describe (or imagine) some examples?
Delete“It occurs to me that maybe we’re heavily invested in finding answers to which we don’t know the corresponding questions. Maybe the availability of the new technology is running the science rather than the other way round”
ReplyDelete“It occurs to me that maybe we’re heavily invested in finding answers to which we don’t know the corresponding questions. Maybe the availability of the new technology is running the science rather than the other way round”
I was compelled by many of Fodors points, and I do think it is valuable to question where scientific money and attention is going. This quote jumped out at me because it resonated with me to some extent. At times it seems that psychological research has been done with little practical application, but I think it’s important to understand that the field is still very young. More information is never a bad thing, but perhaps he feels we need to spend more time on the framework surrounding it. It seems that Fodor takes issue with the seemingly “random” way he feels psychological research is conducted. He thinks that many studies have not been properly justified, although he concedes that clinically there is value in studying loci of neural activity so as to treat patients medically. I would be curious to know what he would like to study instead; if he has an alternative way to study the brain and cognition that he feels is more worthwhile. Overall though I found his ideas refreshing to read. It is nice to consider that there are many ways of approaching these questions, and while we have made remarkable discoveries, we can always be questioning our methods in an effort to advance and improve them.
Fodor (a co-author of Pylyshyn's) was a computationalist, but he was very vague about how to go about it. He was mainly a philosopher who wanted to do for semantics what Chomsky did for syntax (Weeks 8 and 9).
DeleteMadelaine, I agree with you that it was very refreshing to read something that takes a different perspective! We seem to be constantly using up money and attention in order to work towards a goal of uncovering the “answers,” but like you said, the field is still young. Modern neuroscience and cognitive science only emerged around the mid-1950s, so there is still a long way to go, both in terms of discoveries to be made and more questions to be asked. A lot of what both you and Fodor have said has led me to two more questions: do we even know what we are looking for? And how will we know that we have found what we are looking for? If we take the hard problem of consciousness as an example, how will we know that we have solved it? Have we not found the solution because our science and technology are not advanced enough yet, or is it right in front of our noses in a way that we, as humans, simply cannot comprehend or process? I would love to hear more of Fodor’s perspective on how we should improve our question-asking abilities. It may be of value to take a break from searching for answers and reflect on how best to refine the questions we should be asking instead.
DeleteWhat are we looking for? The solution to the easy problem (how and why are organisms able to do what they are able to do?) T2-T4 would provide the answer.
DeleteThe problem of explaining how and why organisms feel is much harder. We'll find out why by Week 10. (But I think you're still mixing up the HP with the OMP..._
Fodor presents his somewhat controversial opinions on neuro-imagining research in this paper and, although I agree with some of his points, I have to disagree with others. For starters, he proclaims (about brain imaging) that “Maybe the availability of the new technology is running the science rather that the other way round.” The article, being written at a time where a fair amount of influential neuroscience-related technologies were developed, is critical of how these technologies were mainly used for brain mapping and not for more varied/original and philosophical work. Scientists loved to use these novel machines and the pleasure they derived from using powerful brain imaging technologies definitely motivated the research they did at the time. Being a cognitive science student and not a neuroscience student, I am less drawn to neuro-anotomy and more to how cognitive skills arise from brain activity and how understanding the brain can help us answer existential questions. Because of this, I do sometimes tend to dismiss brain-mapping research, but I do recognize how valuable it is, unlike Fodor in his text. Brain-mapping is important for clinical diagnoses and treatments. It can also serve as a basis for studying modularity and more abstract cognitive phenomena. Reading this paper made me reflect on the various ways cognition is studied and on the divide that exists between certain modes of inquiry and paradigms. For example, Fodor brings up the differences in how empiricists and rationalists perceive brain functions to be located in the brain. This reminded of the debate between the associative and genetic hypotheses for mirror neurons and of how divisive scientific theories can be. Perhaps we should be critical of how new technologies direct scientific enquiry and may leave aside other theories and experiments that may not seem as novel and fresh, but are equally as valid.
ReplyDeleteScientific disagreement can be useful, even creative, as long as it remains answerable to objective evidence.
DeleteIn some respects, I think I agree with Fodor that knowing exactly where in the brain mental states happen is not the most useful. I was thinking about a study where brain scans were taken of Korean adoptees (who did not speak Korean) and fluent Korean speakers when listening to Korean speech. Despite not knowing Korean, the adoptees brain imaging showed similar patterns to the fluent speakers, presumably due to their early exposure at birth. In essence, the Korean adoptees’ brains showed unconscious recognition despite the fact that the adoptees themselves could not explicitly understand Korean. Although this is interesting, this is a clear example of how knowing where certain processes occur in the brain does not necessarily indicate understanding. That being said, I think there are still benefits from localizing mental states for medical use, just not necessarily philosophical theories.
ReplyDeleteFodor agrees about medical use.
DeleteMarisa your comment sums up how I relate to the article as well, the same areas can light up and yet a major difference between two people is that one understands while the other doesn't
DeleteWould you say that that even if every relevant neuron could be imaged? (That would sound like spooky dualism...)
DeleteFodor comes back to the same issue that we have been discussing; how to get a better understanding of consciousness, and cognition. There are currently many approaches that exists, whether it be cognitive neuroscientists who look at the physical brain, its neural activity and localization of function, or computationalists who believe that everything will be understood algorithmically. Fodor argues that neuroscience helps provide correlation between brain activity, the when and where things happen, and our cognitive abilities, T3, however it doesn’t explain the how or why we are able to do such thing. The brain receives input from the world, and sensory organs, yet the process to generate its outputs, which is everything we do; T3, remains unclear. If computationalism were correct, then looking at the brain, and its neural firing, its hardware is of no use, yet a computationalist approach alone still doesn’t answer questions of cognition, and how it feels like to do certain things.
ReplyDeleteCorrelations between brain activity and behavior are not T3, they are just correlations, whether or not computationalism is correct. What is T3? and T4?
Delete“ But why (unless you’re thinking of having it taken out) does it matter where in the engine the carburettor is?”
ReplyDeleteJerry Fodor’s car analogy for the brain helped me understand why he’s saying that the question of location in the brain is not the most useful one. From what I understood, I see Jerry fordor’s arguments as a way to show us that to understand the brain and the mind, we should focus on understanding their components first without focusing as much on where they are in the brain. consequently, he wants us to focus on function instead of location when it comes to understanding the brain and the mind.
How does focusing on "components" ("modularity") help you figure out how to pass T3 (or T2 or T4). What is "function"?
DeleteRyan, it’s interesting, for me this example almost weakened Fodor’s point. What he seems to be saying is, if what we care about is how something works, why does it matter if we have a picture of what it looks like (where things are). In a straightforward mechanical example, though, to me it seems like where things are can help us understand how whatever is happening is happening. For example, think of a pinball machine where a ball is dropped from the top, moves through a series of paths and levers within an enclosed box, and eventually falls into a slot at the base of the machine where it turns a switch to light up a bulb. If I were wondering how the lightbulb at the end lit up (and why that specific lightbulb lit up rather than the ones beside it), it would certainly be helpful if the front of the box were plexiglass and I could see that there is a ball moving through that eventually turns the light on, and that the location of certain paths+levers determines which lightbulb is turned on at the end. In a simple mechanical system like this, it seem like the how becomes self-evident once we have a complete picture of the system. Similarly, I could see how understanding where everything is in an engine might help us understand how it works, as it could give insight into the causal relationships between parts (at the most minimal level, because for one part to effect another directly they must be have physical contact). All that said, I agree with Fodor that neuroimaging is probably not getting us anywhere on the how question. But to me it seems like this is because the “let’s see what it looks like” approach in neuroimaging is not giving us nearly a detailed enough map for there to be any insight gained from it. It would be one thing if we could map out every single neuron in the brain and track which is activate when, maybe then if would become self evident how cognitive processes occur, but we are so so far from being able to understand at a minute level the highly complex processes that make up cognition, that looking at what area of the brain has activity correlated with a specific capacity feels about as pointless as answering the “how and why does that lightbulb light up” question by pointing out that every time that lightbulb lights up we can hear a ball moving through the pinball box.
DeleteRemember the example in Week 1 about the difference between reverse-engineering the heart and reverse-engineering the brain?
DeleteIn the case of vegetative function, the organ wears its function "on its sleeve": what it does on the outside "resembles" what it does on the inside. You can "see" how the heart does what it does: pump blood.
But you can't "see" how the brain does what it does: "pump" behavior (which is everything that organisms can do: T3, or even T4). There's no resemblance.
If cars grew on trees (rather than being devices we invented, designed and built) a possible rebuttal to Fodor would be that if we had "auto-imaging" that was close to being as complete as what you imagined for the brain-imaging of every one of its neurons and neurotransmitters), then we could eventually figure out how a car works, just as we did with the heart. Same for the lungs, or the kidneys, and possibly even the "vegetative" functions of the brain, such as balance, temperature regulation (homeostasis) and breathing.
But does that work for cognitive function (T3)?
[And remember that when/where imaging would not even work even work if computationalism were true, and the brain were just a computer, i.e., hardware executing software: manipulating symbols. "Cyber-imaging" that showed us what every bit of the hardware was doing at every moment would not tell us what program it was executing (except for the simplest "toy" computer that was only doing what a desk-calculator does).]
To answer prof Harnad's question, I don't think mapping out the brain would work to understand cognitive functions (in the same we we can understand how a hypothetical tree-growing car would works by auto-mapping it's engine). The reason for this is that despite both of them being physical objects, brains use their physical components to represent things symbolically whilst car engines use their physical components only to execute mechanical actions, (with no symbolic representations involved). As a result, mapping both objects would allow us to understand how their different components can physically interact with each other under different conditions. For cars, this means that everything from acceleration to turning can be explained by external forces acting upon the car's physical parts. Similary, interactions between a brain's physcial parts (ex. action potentials between neurons) can help explain how we can do what we do, which is the easy problem. On the other hand, it can't solve the hard problem of how and why we feel because mapping out the brain doesn't indicate how meaning can arise from interactions between it's physical parts, as with prof Harnad's cyber-imaging example above.
DeleteI agree with the existence of the enormous gap that Fordo points out in this article: the gap between the answer of the easy problem, where and when these neurons perform their cognitive capacity, that is, localization of brain functions; and the answer to why, by performing these tasks, the mind generate feelings or "thinking." However, despite his arguments that it is ludicrous to see function localization as the Archimedean point to solve the hard problem of cognitive science, I was expecting Fordo to propose his opinion on how we should reconsider the paradigm of localization of brain function, eschewing the equivocal or "dualism," leveling up the correlation between the activity of particular brain sections and behavior to a causal explanation to the human brain. When it comes to the last paragraph, his argument sounds even more inscrutable to me. He asks "how the neuroscientists decided which experimental tasks it would be interesting to make brain maps for?" I suppose that's also what these neuroscientists ask themselves when they are trying to answer what is going on in our brain. They keep reminding themselves studying only feathers is not enough to understand bird flight; knowing "where" and "when" these neurons are active is not enough to reveal cognition.
ReplyDeleteWhat Fodor says about where/when is already true of trying to reverse-engineer the "easy" problem of what organisms can do, without even trying to solve the "hard" problem of feeling.
DeleteAnd the metaphysical problem of "dualism" ("are there two kinds of 'stuff': 'mental' and 'physical'? i.e., stuff that is felt and stuff that isn't") has absolutely nothing to do with it -- other than noting that solving the "hard" problem is indeed hard. (Why?)
Why do we care? Well, Fodor isn't sincerely asking why in the article, but it made me sincerely reflect on the 'why'. Humans are extremely interested in knowing about themselves. Neurocentrism, I would say, arises from the same collective desire as origin stories. Of course, his critique is on all that going on in studies of localization of function, not all studies of the brain. Does it matter that "now we know it scientifically"? Objectivity, which others mentioned in their comments, is one factor. I wonder if a possible reason is that we want to know what is 'real'. If a plurality of methodologies - brain imaging, introspection, folk psychology, etc.,- converge on the same conclusion, we are more certain of 'the' reality.
ReplyDeleteI can't help pointing out that resources are limited in the music industry which leads to Weber's musicals. The difference is that science is more funded by the public (while artists depend more on how much the public likes it) so that could justify why scientists should be more responsible about how they budget human and non-human capital..
Yes, "neurocentrism" can be a bit like horoscopes (and origin myths).
DeleteAnd multiple methods and theories are certainly helpful if they give causal information. But do brain-imaging data give causal information for reverse-engineering T3 capacity? (Ferna has described some potential toy steps Friday.)
The problem of "resources" (both monetary and material) is far, far deeper than budgetary questions in the entertainment industry or even in science. This will be touched upon in Week 11, the last week. For now, consider climate, covid, poverty, and anti-vaxx...
Will be looking forward to Week 11!
DeleteI think this is a really interesting point April, maybe we are all just narcissists at heart! Fodor notes the abundance of brain talk in publications like the New York Times. Maybe so many people seem to prefer to read about functional localization in the brain rather than astrophysics, for example, because the first teaches us something about ourselves--regardless of if it is actually useful or not.
DeleteOrganisms are survival/reproduction machines at heart (and at brain, if they have brains).
DeleteBut I hope that egoism and narcissism can be constrained by culture and laws -- and that, except in conflicts of life-or-death interests, as between obligate carnivores and their prey, most human beings are (potentially) decent rather than psychopaths.
It seems to me that much of Fodor’s argument hinges on the distinction between correctness and usefulness. In clinical neuroscience, or when dealing with people with specific brain-related problems, it will of course be practical to use our best available knowledge about the brain to help cure them (without, say, cutting out a part of their brain apparently essential for reading English). However, when attempting to answer a more profound question, that deals not with the functional properties of an organism (what it can do and say and what parts of its brain it uses to do and say things), but rather the totality of the organism (how does it function, why does it function), our ‘best available knowledge’ seems to fall short. Anyone surprised at this unfortunate tendency should stop for a second to think about what science purports to do, and compare this to what science is ~actually~ doing. If we hold hypothesis testing and confirmation to lead us toward some absolute truth about the functions of the world (be they brain-related or otherwise), then we are unfortunately digging ourselves into a hole… Poking a brain around and seeing what happens to the person inside that brain does nothing to explain how that brain functions, but it can help us to usefully predict what different kinds of brain injuries would lead to different kinds of functional problems. Notice how this does not produce any ~true~ or correct knowledge about the world, but rather just a useful framework by which we can determine what actions need to be taken. Whether scientific study even has the correct tools to get at the actual “how” and “why” should still be up for debate, in my opinion. If “how” and “why” are even answerable questions, that is !
ReplyDeleteI'm not sure I follow you. The distinction between clinical neuroscience and cognitive science (including cognitive neuroscience) seems clear enough.
DeleteWhat is the problem you see with trying to reverse-engineer the cognitive capacities of the brain (apart from the fact that reverse-engineering feeling capacity is going to be much harder than reverse-engineering doing-capacity?
I think my issue boils down to a lack of knowledge, or lack of the right kind of knowledge. If we do intend to reverse-engineer the cognitive capacities of the brain, we will have to do so using the knowledge we currently have. And if our knowledge of the brain is composed mainly or entirely of 'useful' knowledge, that is, functional knowledge we have garnered through testing hypotheses about how the brain works, then we will be building a system that is reverse-engineered in the vein of "good enough". But I'm not sure that that's what reverse-engineering really is. Shouldn't we have to start from some fundamental or 'correct' knowledge about how/why the brain works, rather than just some integral of everything we sort-of know about what the brain does? Would it not require some breakthrough or strategy-shift in both clinical and cognitive neuroscience in order to reach this kind of knowledge? I am not sure what this would look like, which is what breeds my skepticism as to our ability to ever actually reverse-engineer something that will pass as human.
DeleteWhen Fodor discusses the question of whether mental functions are neurally localized in the brain and whether some mental processes are “sui genesis”, this reminded me of the discussion we had in class about T2 and T3. As “Stevan says”, in order to pass T2, you need a T3 robot. If some of the mind/some mental functions are elsewhere from the brain, this would support the idea that a T3 robot would be needed to pass T2. Connecting this to Cook, et al. (2014), if the associative learning hypothesis is correct, sensorimotor learning is necessary for MN development. Therefore, some cognitive functions arise from the body. For example, your motor neuron axons that run down the arm are involved in the process of recognizing and imitating another person bringing food to their mouth. That would mean the mental function doesn’t lie solely in the brain.
ReplyDeleteWell, the PNS and CNS are both neural...
DeleteWhat do you mean by "mental"?
I think that Persinger's God Helmet is a thoroughly apt example for exploring Fodor further. For those who aren’t familiar, inspiration from Temporal Lobe Epilepsy patients led Michael Persinger to create the God Helmet. Temporal Lobe Epilepsy patients often reported intense experiences of godliness, religious upheavals, sensed presences, and feelings of being ascended to a spiritual realm. Through these testimonies, Persinger set out to prove that the temporal lobes held the hard-wired circuitry for religious experiences. He made an instrument which electrically stimulated this area, and then he asked participants to explain their experience and internal state during the experiment. Many reported those exact feelings of these epilepsy patients, some reporting the closest to their god they had ever been, but most only reporting a sensed presence in the room with them.
ReplyDeleteWhat do we learn from identifying the religious apparatus in the brain? Fodor would probably argue not much. There is such a complexity around human religious experience, to simplify it to the stimulating or firing of a broad brain region does not tell us about why we experience God(s) from this, or even how this neural excitement can lead to such culturally significant traditions around the world. “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.” Maybe neural map making is just another form of scientific stamp collecting, and we need new multidisciplinary projects in order to integrate humanity with its brain.
Good points (but the fact that "religious" experiences can be triggered by certain brain areas, chemicals or epilepsy may have some clinical significance). (And be careful, Fodorian scoffing is contagious and can lead to destructive scepticism!)
DeleteThere's still the kinds of hypothesis-testing experiments described by Ferna, using both behavioral and neural measures. Even when the hypotheses are small (toy) they could grow.
Fodor says in his article that imaging where things occur in the brain would most likely not yield to any significant discoveries about how and why the mind functions because “[...] our intuition gives us enough knowledge to make neuroimaging irrelevant.” So far, our breakthroughs in reverse-engineering the brain appear to have simply taught us what we already knew intuitively. If we're just entering into neuroimaging with the concept of "let's see what we can discover" he claims, that's not a very effective method of conducting science. Conversely, in order for neuroimaging to be effective, some query or hypothesis about the nature of cognition must be addressed or verified. According to him, we can't truly begin to develop meaningful theories of cognitive architecture until we have a clearer understanding of the fundamental nature of cognition. Brain imaging must be done correctly, with the goal of evaluating the validity of complete theories of cognition rather than toy dilemmas.
ReplyDeleteFor an example, see Ferna's McGill IPN doctoral thesis (2019), "How and why does category learning cause categorical perception? Behavioural, neural and computational aspects"
DeleteI think Fodor is right in the sense that it should be considered a waste of resources to conduct experiments with no real purpose. In his concluding paragraph he describes asking neuroscientists “how [they] decide which experimental tasks it would be interesting to make brain maps for” and how this left him wondering if they were really wasting their time. I think this is an overgeneralization however. If this were true, and neuroscientists had no idea what they were looking for, it does not explain the new methods that have been developed that provide a different understanding of brain areas. For example, the use of fluorescent proteins allows for staining of activity as well as pathways. The creation of new imaging methods could only have been provoked by the need to know more than what was currently available and the need to know more could only have been driven by real questions. This goes unmentioned in Fodor’s article and I think it paints neuroscience as a discipline floundering for answers due to the fact he only highlights what is really just bad science.
ReplyDeleteHowever, this shot-in-the-dark method does not have to be framed as completely useless. Lots of scientific discoveries resulted from scientists stumbling upon them. For example, Hans Selye, the founding father of stress research discovered the consequences of chronic stress by noticing that all of his lab mice had developed ulcers after undergoing daily injections. This isn’t me trying to defend bad science, it is merely pointing out that the less specific an experiment is, the higher the potential for unexpected outcomes .
Do you have examples from cogsci (reverse-engineering cognitive capacity) rather than clinical medicine?
DeleteWould an example of this be the active and passive kittens experiment? From what I recall, there was no specific hypothesis laid out they merely wanted to examine what differences would emerge between the kittens. Clearly horrendous ethical problems aside, what they discovered about the kittens following lesioning and exploratory behaviour was not a targeted hypothesis, yet they still came to a scientific conclusion.
DeleteFodor’s main question is why do we need to know “where”, when we already know “how”. He raises questions about the real significance of knowing where things are exactly localized in order for us to understand how. But isn’t the answer to this obvious? Knowing where things are located enables us to understand how things are connected. Understanding how things are connected is essential for understanding how the entire system works, since it’s composed of many parts and not just one. He asks why we need to know where the carburettor is to understanding how the engine works. I think it’s important to note that the information of “where” on its own is not helpful, but when applied contextually to where other parts of the system and their function are located, it is very useful for understanding how parts of the system interact, or what might happen if one of them is removed.
ReplyDeleteIf we already know how, what's left for cogsci to do?
DeleteFodor's saying we don't know how, and brain imaging's when-and-where does not explain how.
Looking at what's connected to what helps us understand how the heart works, because the heart just pumps blood. But the brain does everything we can do: it's not clear how we can "read that off" what's connected to what in the brain the way we can with the heart (or maybe even a car! Fodor's example is a weak one).
Does knowing what's connected to what (and what's active when and where) in the brain help us build (or even model) a T3 robot?
In response to this post, I don’t think that knowing when-and-where would really help us how. As mentioned in the reading, knowing the location of where something happens doesn’t actually give us much information on the functioning or the how behind any action. I also think that the brain is much more complex than to say one thing would have one specific are in the brain. Although regions of the brain may have a specific function, the more complex the task/thought/action that is performed, the more areas of the brain will be engaged.
Deletetaking this complexity into consideration, it just makes it harder and somewhat useless in trying to locate the where instead of trying to understand how it all integrated. Although it may not give us a lot of information on the know, I think that the location of the “where” of certain general areas in the brain may allow us to build a better T3 model. If we are able to have a better model, we may be able to reverse-engineer a better design of the brain allowing us to arrive at a better understanding of the how.
lkj
ReplyDeleteI agree with the analysis already noted by some of my classmates, that the quest for functional localization of mental functions may be related to the ever-present quest for objectivity in science. Fodor's criticism of functional localization reminded me of how past professors have described the human genome project - probably a waste of time and resources. I think that an objective map of mental functions or of all genes in the human genome may be so desirable because a map implies that it can be completed. Thinking in terms of functional localization may be exciting to scientists because it feels as though each discovery is one piece closer to solving the puzzle. However, what if the whole is more than the sum of the parts? I think this relates to the hard problem of consciousness. If neuroscience manages to somehow map every mental function and consciousness can still not be reverse-engineered, where do we go from there?
ReplyDeleteAlso, I am still a bit unsure about the parallel that Fodor draws on the argument between homogenous and heterogeneous minds and that between empiricists and rationalists. I understand the general distinction made in his explanation, but I am confused as to how this specifically relates to empiricism and rationalism, as I understand them.
Fodor's scepticism about whether we can explain how-and-why from when-and-where is already a problem for the easy problem.
DeleteBut forget about Fodor's homogenous vs heterogeneous minds and empiricists vs rationalists -- that's just unilluminating spill-over from philosophers' debates...
Fodor was hoping to find for semantics a “poverty of the stimulus argument” analogous to Chomsky’s “poverty of the stimulus argument” for syntax. Except that Chomsky was right about syntax (as we will discuss in Weeks 8 and 9 about “Universal Grammar”) but Fodor is wrong about semantics: Universal Grammar is unlearnable, so it must be innate. But word meaning is learnable, so it is not innate.
What is the “poverty of the stimulus argument”?
I think it is interesting to see an argument like Fodor's, as it gives a completely different perspective to neuroscience and neuroimaging. However, as he is coming from a philosophical viewpoint, there will be a lot of natural discrepancy from the thoughts of one of the neuroscientists and their experiments which he was skeptical of. I believe the brain-mapping is a natural and essential step in progress towards understanding the brain in more detail, although the general question of whether we can even understand the "how and why" is rightly debated on the philosophical level. Looking through history, scientific study was always required to advance medical knowledge, and philosophical skepticism was not usually as important to that progress, although it does help guide and scrutinize the science that is being studied at that point in history.
ReplyDeleteAlthough I found myself agreeing with most of Fodor’s points, I don’t agree with his cynicism with respect to experimental data. While I understand this perspective, I think that artificial intelligence has changed the game when it comes to data, even if this data was not collected in an ideal manner and may seem insignificant. This is because, in recent years, certain AI has been developed that can take it massive amounts of data and find patterns in it that humans would have been unlikely to find. By AI, I don’t necessarily mean a Turing Machine geared at understanding the easy and hard problems, as we’ve discussed in lectures. I mean specialized AI geared towards a specific purpose. AI keeps getting more and more sophisticated extremely quickly, and it doesn’t seem like this trend will stop anytime soon. It is not unreasonable to think that including experimental data from brain localization studies in the data fed to a certain program could yield unexpected results and, potentially, advance our understanding of the brain. I’m not saying that this is likely, but it is certainly more feasible that it was when Fodor wrote this article in 1999.
ReplyDeleteI definitely agree that simply mapping out brain activation differences is completely useless and that we should move out of this trend in neuroscience. Brain imaging is a tool (a computational tool) that can be used to test some hypothesis. Just like how we can't understand cognition simply with computation, we can't do it with brain imagery.
ReplyDeleteThe only saving grace for this trend is in my opinion all the advancement we have made in brain imagery in the past decades.
All those expensive research projects have probably served to advance technology and statistical methods for computing useful results in clinical neuroscience using brain imagery. Sometimes you need to focus on making more useful tools, before you can even start to address the more important questions (i.e. how does the brain work?) -Elyass
I think Fodor's argument in this article is twofold: he is arguing against the preoccupation with brain localization and ultimately against the tendency to take neuroimaging data alone as an explanation of mind.
ReplyDeleteI like his analogy between mind and automobile. It articulates his point quite well. "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?"
I think the point Fodor tries to make here is that what is important is at the functional organizational level, the implementation details are not that important: what we want to know is how cognitive functions are organized in a structure; whether a function is implemented in a particular location or not does not matter really.
I think his anecdote of Pavlov is a little bit misleading, because it seems to me he is trying to argue against scientific methodology or question the value of scientific experiments and data. But from reading the whole article, I found his view is exactly the opposite. He is actually criticizing those who obsess themselves with brain imaging for being not scientific enough. The ultimate point he wants to reach through his anti-brain-localization argument is that we should not place mere scientific data over a scientific theory. Brain imaging data should be used to test theories of mind, and a theory of mind should explain things at the functional level. Simply mapping brain regions does not verify or falsify any theory, since by his automobile analogy, it is irrelevant to what happens at the functional level.
What I am wondering is that: do we even figure out and have a consensus on what neuroimaging data counts as evidence for a theory at the functional level and what does not count? Maybe before we can do that, the scientific community need first to agree on how these two levels are related, which requires a wide accepted theory on this issue in the first place.
My original comment/skywriting for 4B was somehow deleted so I am reposting:
ReplyDeleteAfter reading Fodor’s article last week, my first initial thought was “Yes, Fodor’s point makes sense”. My second thought/emotion was confusion. Not in the sense that I was confused about the irrelevance of neural correlates in explaining the how mechanisms of mirroring capacities but a confusion about the field of neuroscience and its connection to cognitive science in general.
The question that kept whirling inside my head was: “Surely, depending on which body of philosophy you are based on, (empiricism, rationalism, dualism, materialism, functionalism etc.), the weight you put in the significance of studying the brain must differ but regardless of philosophy, neuroscience must have some value to cognitive science? No?”
Fodor’s article was very explicitly implying “No, neural measures and localization of function is not doing much for cogsci.” And as Prof Harnad mentions above, Fodorian scoffing is indeed contagious and leads to destructive skepticism. I needed to distance myself from such scoffing. Fernanda’s talk and the experiments she introduced us to has helped recalibrate my thoughts.
Science has trends. In the field of cogsci, it seems that the early trend (other than introspection and homunculus) was behaviorism. When behaviorism was the trend, academia was over thrilled with hypothesis-testing experiments. Experimental psychologists studied the brain’s capacities through behavior, resulting in immense behavioral data. Then came the cognitive era along with computationalism, and neuroimaging rising to the surface. Again, neuroimaging allows for hypothesis-testing experiments which leads to the situation Fodor describes in the article. “Availability of the new technology is running the science rather than the other way around.”
All these experiments leave us with mere neural correlates of cognitive capacities. Both behaviorism and neuroimaging have clinical uses (hooray for that) but not so much when it comes to reverse engineering cognitive capacities. I think this has been made clear by Fodor and Prof Harnad but Fernanda’s perspective shows us some value of neuroscience.
She did mention that neuroscience has started moving away from function localization and is now trying to understanding function through complex neural networks. She also mentioned that neuroscientists are combining neuroimaging techniques with computational modeling to test mechanistic hypothesis which seem a lot more purposeful. As Prof Harnad points out, these experiments hypothesize about toys, (not T2 or T3) but could work as steps.
Sorry about the vanished posting. (Probably my fault when I permanently delete skies that their authors have removed.)
DeleteYes, neural measures can be used along with behavioral measures in testing causal models (along the road to T3 or T4).
Many critics towards neuroscience are about reductionism. As an empirical science study, it relies on evidence like neuroimaging but at the meantime, we should be aware that activation in brain areas can only be (or sometimes even cannot be) “correlates” of mental process. I realized in recent year neuroscientists are proposing studies that follow the formula of “investigating the role of xxx(brain area) in the xxx(task)” and then conducting neuroimaging experiment (often in a small number of participants, considering the high costs of fMRI). However, the behavioral tasks humans are capable doing is infinite, and these kinds of experiments are also infinite. Empirically speaking, sometime we couldn’t even know what brain activities we are recording (e.g. if an experiment task for participants is to “relax” or to “imagine xxx”, how you make sure they’re not mind-wandering, or they are actually following the task?). We must be critical about experiment like this and think from a cognitive science perspective that – can functional localization (when and where) tell us how brain/cognition work? I suppose not. Therefore, I believe neuroscience can only be one of the methods for studying cognition(that is, yes, I still care about the neuroscience), but try to avoid reductionism and always be open to cognitive science studies from other disciplines.
ReplyDeleteWhy does everyone go on so about the brain? I guess maybe the obsession of understanding every part of our body, every function of each part. When we are trying to construct a T3/T4 to understand human cognition, I feel like we are trying to construct a new us. Humans are always bold and in quest. That is maybe why, understanding brain, this mysterious organism which controls our body, is this attracting. (But I agree that neuroimaging offers little help with How)
ReplyDelete“Although Turing is surely right about the universal power of computation, it does not follow that all there is in the world is computation […] They can all be simulated, computationally, by computation. But the simulations do not have the electrical, chemical, and other dynamic properties of the real thing.” This quote from “Can neuroimaging reveal how the brain thinks” makes me questions: by definition, T3 is indistinguishable in sensorimotor performance capacity, T4 is indistinguishable in both external performance and internal function, are they considered as a simulation of human? If T4 is also indistinguishable in internal function, T4 is no longer a simulation for also having same properties as humans. Then, how do we classify them?
When Fernanda explained in class the research being conducted regarding mirror neurons, it became even clearer to me that correlation is not causation. Just because an area ‘lights up’ in an fMRI, doesn’t explain much as to why it lights up and how it helps in the cognitive function. For example, the study that ‘located’ the presence of words throughout the brain was not accurate in its conclusion, because it just tried say something like ‘oh now we know what regions of the brain are involved when you’re thinking about a particular word’, which from what I understand, is wrong, because the regions of the brain are so interconnected that saying that one particular brain region is involved in something is to ignore the role of other regions in that function.
ReplyDeleteI really enjoyed this reading. Maybe it was the casual blog post style and not dense and convoluted academic writing. But I think Fodor’s frustrations are much like frustrations seen with the studies of “Ethical AI”. Particularly, when Fodor menitons “Science is expensive, and it’s largely publicly funded, and there’s never enough money to do all the research that might be worth doing.” There is the investment of valuable attention from leaders in their fields of computer science, psychology, philosophy etc, and millions, perhaps billions of dollars if we count all the ethical AI research teams across institutions including big tech, universities and other independent AI research institutions. Their primary concern is how we as humans will interact with the AI, and whether the rules/laws/regulations that govern us will extend to them as well. Personally, I am neutral on this topic, however I understand frustrations arise as there are sentient beings that are being overlooked in this process. I think what interests me most about connecting this reading with publicly stated sentiments of lack of care for Ethical AI discussions. This has piqued my interest in the definitive hierarchy or prioritization of research objectives in academia and industry. Perhaps they are similar to some, but I think the former is grant focused so highly dependent on the funding institution and the latter has business objectives that translate to more applied work or at least more apparently applied work. So how can we best distribute the resources to focus on the right problems? The scope I’ve touched upon is computationally focused and somewhat relevant to this class, but this extends to many of the “blue sky” problems in research in other domains as well.
ReplyDeleteAnother point that came up in this article was the idea of empiricism versus rationalism. More specifically, the idea of localisation of function in the brain versus the equipotential camp. I think it’s interesting that in 1999, fodor believed that the rationalists who thought there “might be something to phrenology” lost the battle against empiricists but that they will win the war. I’m generally curious as to what the widespread beliefs of philosophers/neuroscientists at that point considering some things that were already known about both at the time this article was published. For example, phrenology was largely discredited in the 1840s, and by the 1990’s, Broadmann’s popularized areas of the brain was largely accepted. Why did Fodor think that phrenology would eventually “win the war”? Maybe this was a specific example I’m reading too closely into, but I’m curious as to his reasoning.
ReplyDelete