4a. Rizzolatti, G., & Destro, M. F. (2008). Mirror neurons and their function in cognitively understood action
Why is there controversy over whether neuroscience is relevant to explaining cognition? (We could figure out how the heart can do what it can do: pump blood. But the brain can do anything and everything we can do. That's what it pumps.)
Reading: Rizzolatti, G., & Destro, M. F. (2008). Mirror neurons. Scholarpedia, 3(1), 2055.
Optional:
Bandera, J. P., Marfil, R., Molina-Tanco, L., Rodriguez, J. A., Bandera, A., & Sandoval, F. (2007). Robot learning by active imitation. INTECH Open Access Publisher.
Cook, R., Bird, G., Catmur, C., Press, C., & Heyes, C. (2014). Mirror neurons: from origin to function. Behavioral and Brain Sciences, 37(02), 177-192.
Bonini, L., Rotunno, C., Arcuri, E., & Gallese, V. (2022). Mirror neurons 30 years later: implications and applications. Trends in Cognitive Sciences.
It seems like the view of cognition as solely being a product of computationalism causes some to think the mechanisms of the brain do not matter for clear comprehension of thinking. Studying the brain is much harder than studying the heart because it is multifunctional and, therefore, not as worthy of study due to the high difficulty involved. I personally disagree with this, as computation is highly unlikely to be the sole explanation for consciousness (as seen in the previous reading with Harnad, which states that computation is not sufficient to explain understanding). We can gain deeper insight into cognition through studying the brain’s structure despite its complexities. For example, the current reading helps to explain why autistic people have difficulties with understanding intentions through a dysfunctional mirror neuron system. This would be hard to explain solely through computation. Studies of specific components of the brain, such as mirror neurons, bring insight into various aspects of our capabilities, including the origins of language, empathy, action imitation, and more.
ReplyDeleteI agree with you (and partially with Searle) that the brain should not be ignored when studying cognition, especially after reading about mirror neurons. This reinforces Dr. Harnad’s point in reading 3b on redefining Searle’s second tenet from “the brain is irrelevant” to “computation is implementation-independent”. While I agree that the brain is relevant, I believe that any dynamical system could eventually perform the same computation that explain cognition. I will also use the example of differences of mirror neuron system in autistic individuals. Working with autistic children, I developed a research interest in understanding why many of them do not develop speech for functional communication. This could not only help these children develop alternative ways of communication (I do not agree that sign languages are not “natural human communication system”), but also answer the question about the evolution of language. As I am reading more about computation for this class and how it can be a scientific method to explain causal mechanisms of cognition, I am wondering whether a T3 or T4 machine could explain causal mechanisms of speech difficulties in autistic individuals.
Delete
DeleteIsabelle, it is not that the brain is less worthy of study! It's just much more complicated, because it does so many things. (And see 4b, Fodor, on how "where" and "when" in the brain cannot explain how "how" or "why" we think.) But it's true that computation alone cannot explain it either.
A "dysfunctional mirror system" may explain dysfunctional behaviour, but CogSci needs to reverse-engineer mirror capacities to explain how it produces them when it is functional! But, you are right that in the process of scaling up to full T2/T3/T4 capacity, mirror-capacities in "language, empathy, action imitation, and more" will need to be reverse-engineered...
Anne-Sophie it's not that "any dynamical system could eventually perform the same computation that explains cognition." Maybe more than one, maybe even many, but certainly not any!
You are right, though, that "sign languages" are “natural human communication system” -- perhaps reverse-engineering them will help cogsci to understand the evolutionary origin of language (Week 7).
You ask "whether a T3 or T4 machine could explain causal mechanisms of speech difficulties in autistic individuals." But first they have to successfully reverse-engineer and thereby explain neurotypical language capacity...
***EVERYBODY PLEASE NOTE: I REDUCED THE MINIMUM NUMBER OF SKYWRITINGS. BUT THE READINGS ARE **ALL** RELEVANT TO AN OVERALL UNDERSTANDING OF THE COURSE. SO, EVEN IF YOU DO NOT DO A SKYWRITING ON ALL OF THEM, AT LEAST FEED EACH READING YOU DO NOT READ TO CHATGPT AND ASK IT FOR A SUMMARY, SO YOU KNOW WHAT THE READING SAID — OTHERWISE YOU WILL NOT HAVE A COMPLETE GRASP OF THE COURSE TO INTEGRATE AND INTERCONNECT FOR THE FINAL EXAM.***
DeleteThe article discuss the role of mirror neurons in action understanding and social cognition, suggesting that these neurons form a basis for empathy and even indirect understanding of others’ intentions. While mirror neurons clearly show how observed actions can be internally simulated, there seems to be a gap between this neural mechanism and more complex cognitive phenomena like reasoning, abstract thought, or language. For example, simulating someone reaching for a cup doesn’t necessarily let you understand why they want the cup, or allow you to infer abstract concepts beyond observable behavior. Similarly, if we think about AI agent(a system could have a mechanism analogous to mirror neurons) that can activate internal representations of observed actions, but this wouldn’t guarantee human-like understanding or intentionality. In terms of the TT, an AI might imitate goal-directed behaviors convincingly, but it might lack the underlying experiential or affective states that humans attribute to those actions. This suggests that mirror neuron mechanisms might be necessary but not sufficient for fully human-like cognition, particularly in social and linguistic contexts.
ReplyDeleteI would agree that no single neuron can explain the complete continuum of human cognition, yet as you stated, it does represent an early and possible important experience of bridging action and understanding. Rizzolatti & Destro (2008) outline that the mirror neurons fire both during action and during experiencing someone else acting, and can potentially serve as a scaffolding for empathizing, imitating, and decoding learning. Further studies on mirror neurons such a s Bonini et al. (2022) continue to show a profound connection to socially mediated cognition yet Cook et al. (2014) also caution pushing the findings too far. It's also true that when you see a person grab a cup, you may not know the precise reason they wanted that cup, but some form of neural simulation occurs, and that sparks an inference you can formulate regarding intentions and goals. Bandera et al. (2007) in robotics also display a factor of social mediation in imitation learning where the robots exhibit imitation learning of action processes, although as you stated, that works on the basis of building connections, not understanding. The overlap is that the mirror neuron theory is an application that helps put together the pieces of the puzzle; however, we won't claim they are the whole equals all of cognition, but potentially they play a role.
DeleteI responded to Rachael H. 's discussion but upon further thought I wanted to say that I agree that mirror neurons don’t explain all of cognition, but I think their value lies in how they provide a starting point for social inference. As Rizzolatti & Destro show, the same firing occurs when we act and when we watch others act, which gives us a mechanism for approximating others’ intentions. While this doesn’t solve the hard problem since we can’t directly access another’s subjective state, I do believe it does help explain how we can bridge the gap between observable action and feeling/meaningful understanding.
DeleteRachel, you are that MNs don’t explain mirror-capacities; they just correlate with them. We are waiting to hear what the causal mechanism is. So they don’t explain any other cognitive capacity either. But although we knew had mirror-capacities (so the brain could produce, somehow) the MN work (and speculations) did draw attention to a number of perhaps relate analog capacities of humans and other species too.
DeleteNeither humans, nor other species, nor evolution itself can “mind-read” (because of the impenetrable Other-Minds barrier.” Yet the capacity mind-read could nevertheless evolve: How?
And what about the power of language to “mind-write”?
Lorena, you are right that although the discovery of the existence of MNs does not explain, it does bring together many similar and perhaps related mirror-capacities.
(What is a mirror-capacity? Give as many examples as you can? Might understanding language also be a mirror capacity? Not just speaking it, but understanding it?)
Sannah, you are right that MNs do not explain all of cognition: they don’t explain anything (even the mechanism producing mirror-capacity!). But in drawing attention to a variety of mirror-capacities, the discovery of MNs does teach us something: What?
(If they cast light on any Problem, it’s not the HP but the O-MP: Why?)
Thanks for the response! As best as I can understand, a mirror-capacity is any form of capacity where the same circuitry in the brain is firing, both when we do something and when we see someone else do it. Rizzolatti and Destro (2008) gave the classic example of grasping: if I reach for a cup, the same neurons that fire in my brain are also firing when I watch (and in that case the same pattern is evidenced) another person reach for a cup. Other examples such as facial responses, like a smile or frown, where simply seeing another person smile or frown (could) activate the same motoric patterns in me (Cook et al., 2014). Perhaps even empathy could be associated with this, in that when I empathize with another person's sadness, similar circuitries (Bonini et al., 2022) could be activated as used when I feel sad. Even in imitation learning, I believe this is likely a function of a mirroring process as well with children or with robots (Bandera et al, 2007). I also find your question thought-provoking with respect to language, because is understanding spoken language and listening perhaps at least partially a form of mirroring capacity? Would my brain trigger a similar brain circuitry that comes into play when I am the one speaking?
DeleteLorena, good summary, except it's not just seeing! Mirroring can occur in any sensory-motor modality in which there is a perception/production analog relation between the "shape" of sensory input and the "shape" of motor output. The most prominent example is vocal imitation, which does not depend on vision, and occurs in humans as surely as it occurs in birds, both within and between species. It has even given rise to a "motor theory of speech perception": Motor Theory of Speech Perception and Mirror Neurons: A Review
DeleteI think one of the most important details from this paper is this predictive property of mirror neurons. These cells aren’t only responding to observed movements; they're firing in anticipation of some predicted outcome, even before the event begins. These cells are engaging in active prediction, not passive imitation, and this ability seems to be founded in sensorimotor memory, which allows for the prediction of future states based on the recall of similar past experiences.
ReplyDeleteA strict T2 system shuffles outputs to match inputs, lacking the ability to ground symbols in sensorimotor experience. In theory, a T3 system would over-come this by interacting with the world through sensorimotor faculties (eg moving, seeing, etc). But mirror neurons raise the question: is anticipation itself a necessary condition for truly passing T3? If so, then the concept of memory is brought into the discourse. Passing T3 would require not only the ability to connect symbols to experiences in the real-world, but the ability to store/recall past interactions to predict what is to come. Without memory capabilities, wouldn’t passing T3 merely be a (superficially) more complex version of T2 that is just able to deal with more intricate input-output patterns, without progression towards genuine understanding?
I find it really interesting that you highlighted the anticipation aspect of mirror neurons. The idea that these cells engage in active prediction, rather than just passive imitation, opens up so many questions about how we connect symbols to real-world experiences. The paper mentions that it’s not fully clear which neurons mirror which actions, or how precise or broad that mapping is, which makes me wonder if there’s more to mirror neurons biologically that we haven’t explored yet. Perhaps this could give us deeper insight into how humans predict and actively engage with the world. I also wonder if understanding this better, maybe in connection with systems like the default mode network—how our brains continue processing even when we’re not actively aware—could inform AI design, helping create systems that actually approach T3 by grounding symbols, using memory, and predicting future states rather than just responding to inputs.
DeleteWhat I found interesting is how this prediction was experimentally confirmed. In the monkey studies, that same initial movement (grasping) triggered different neural responses depending on whether it was part of the grasp-to-eat or grasp-to-place intention. These neurons clearly were not just mirroring the action, but were coding for the next step in the sequence. As you discussed, T2 systems can adapt to input-output patterns through learning, which may resemble anticipation (as highlighted in The Chinese Room Argument), but it is entirely pattern-based and not grounded in real-world meaning. T3 systems require anticipation that is meaningful and contextually grounded. The prediction is based on the significance of the actions themselves. Mirror neurons provide a hint that this anticipation may be what differentiates genuine cognition from highly sophisticated symbol manipulation.
DeleteElle, good point about mirroring and anticipation.
DeleteBy “T2” I think you meant C=C computationalism. But T3 (lifelong verbal and sensorimotor capacity Turing-Indistinguishable from our own) would surely demand the capacity for both imitation and anticipation, wouldn’t it? And memory… You seem to be underestimating T3.
Rachel H, good points (but aren’t they still mostly speculations inspired by mirror-capacities than causal mechanisms that explain them?)
Emily, what passing T3 needs is the causal mechanism of cognitive capacities indistinguishable from our own. Hold onto that criterion; it cannot let you down. Computation (according the the Strong C-TThesis) should be able to come up with a recipe that is interpretable as the solution, 3D-printing (with real robotics) is needed to test whether it really works, or just on paper.
I find it fascinating how mirror neurons give us an almost direct window into how others might be feeling, which relates to my question about AI’s potential for empathy compared to our own. By allowing us to imitate others’ actions and predict their next moves, these neurons provide a biological mechanism for connecting through empathy and simulation. Still, this doesn’t fully solve the problem of understanding meaning or truly knowing what another person is experiencing—because we’re never seeing the world directly through their eyes. This is somewhat reminiscent of AI systems like ChatGPT, which imitate patterns and predict responses based on the 'Big Gulp'. Perhaps this mirroring is as close as humans can biologically get to “walking in someone else’s shoes.” I wonder whether it might be possible to reverse-engineer the functioning of mirror neurons and apply this understanding to AI through cognitive science, to better explain why this process occurs and potentially enhance machine empathy. While mirror neurons give us a practical, predictive insight into others’ minds, each of us has different upbringings, perspectives, experiences, and brains, meaning we can never fully know what someone else is feeling. Mirror neurons may take us as close as biologically possible—but not all the way.
ReplyDeleteYour response made me think of this sentence from the reading: “Recent evidence suggests that the mirror mechanism is also involved in empathy, that is in the capacity of feeling the same emotions that others feel.” While I agree that mirror neurons only give predictive insight into others’ minds, because the mirror mechanism lets us feel the same emotions, can’t we approximate equivalent experiences through inference? Of course, inference can’t give certainty, like the example from class that if my tooth hurts, I can doubt that I have a tooth and that the pain I feel is from a toothache, but I cannot doubt that I feel the pain, but doesn’t this inference still bring us closer to bridging the gap between simulation and feeling?
DeleteRachel H, we can’t mind-read, but we can ∫“mind-write” through language: tell others what we feel. If I understand what “I am tried” means, then I know what it means when you say it. Language production/understanding is a mirror-capacity.
DeleteThe foremost goal of cogsci is to reverse-engineer and test the causal mechanism producing human cognitive capacity. Applying that to make AI more useful for is secondary.
But human mirror-capacity has not been reverse-engineered yet, so nothing to apply,
Sannah, yes, inference can complement mirroring (just as verbal description can), but it’s not as good as being able to mirror the experience itself.
But about certainty: forget it. The only feelings you can feel are your own.
Sannah, I agree when you say that "mirror neurons only give predictive insight into others' mind" but I disagree with the explanation you give. You say it "let us feel the same emotions", but I believe this assumes a lot, i.e. too much, considering the Other minds problem. I would argue that mirror neurons allow us to "project" what our own emotions would be, and we infer, subconsciously, that it MUST be what this other person is feeling. Now to remove the WWs from that, what I mean is : mirror neurons let us feel the emotions we expect the other person to be feeling, given that those emotions would be the ones we would feel in their given situation. I think put this way it reduces the assumptions and "magical capacities" we seem to be giving to mirror neurons. However I do agree with you that our approximation of their feelings might be faulty due to the differences in experiences.
DeleteTo link it to Professor Harnard's point, I would say that mirror neurons allow us to project our potential emotions, hence giving us an insight on what the other person might be feeling, but it is only with "mind-writing" that we are able to know what the others really feel.
I’ve always loved the idea of mirror neurons. I first learned about them back in CEGEP during an anthropology class, and what stuck with me is how they show that the mind is never just passively observing. It’s always leaning into the future, trying to anticipate someone’s next move. That predictive quality makes social interaction feel less like reaction and more like this constant, silent choreography between people.
ReplyDeleteAzul, yes, but how (and why) do MNs do what they do?
DeleteI think the “how” has to do with the way mirror neurons link what we see with our own motor system, so watching someone move activates a kind of internal copy in us. That could explain the “why” too, it could have evolved to make imitation and social learning faster, since we don’t just observe but are already primed to act. But I like how you, Azul, described it as a choreography, because that captures the feeling of it super well!
DeleteRena, yes, and now it's time to do some toy robotic designing and testing to produce mirror capacities.
DeleteThis paper was very interesting as I find mirror neurons so fascinating. However, after the first two sections in sounded like somewhat of a repetition of some previous concepts. The idea that a machine is just mimicking what a human does, resembles this idea that mirror neurons may aid in mimicking and from there, help us reproduce the behaviour. They are both a neural correlates because there is no proof of not subjective experience. Behavioural equivalence doesn’t guarantee genuine consciousness. Similarily, mirror-neuron activation doesn’t prove we literally feel what the other person.
ReplyDeleteKaelyn, I agree. It seems as if we are jumping the gun on mirror mechanism and its link to intentionality. There's a risk that neuroscience is just identifying a correlation and not a causal explanation— which is probably why some scientists wonder whether neuroscience is relevant in explaining cognition because cognition has to do with the things we cannot empirically see (like thinking). However, in order to reverse engineer cognitive capacities through the scientific method, we must start with these correlations as they are a form of empirical evidence. What I think could be missing is evidence on how the mirror mechanism integrates with higher-order systems for context, memory, and prediction. Until neuroscience captures that integration, we may still have a system that simulates behaviour without inherent understanding, like the machine that can mimick but not experience.
DeleteKaelyn, I tend to agree with how you connected mirror neurons to the idea of mimicry, it’s a good reminder that just because a system can reproduce behavior, that doesn’t necessarily prove it feels or understands. And Maya, I think your point about correlation versus causation really gets at why neuroscience’s role in explaining cognition is controversial. Mirror neurons clearly map actions to motor counterparts, but that’s not the same as showing how subjective experience arises. So maybe the tension is that neuroscience gives us powerful clues about how capacities are carried out, but it doesn’t necessarily resolve the deeper “what it feels like” side of cognition
DeleteKaelyn, mirror-neurons (MNs) cannot penetrate the Other-Minds Barrier: They are just mirrors, not periscopes. Nor do they solve the Hard Problem. Turing's Method remains behavioural, observational. Not even Darwinian evolution (Dawkins's "Blind Watchmaker") can mind-read (let alone feel the feelings of others).
DeleteAnd yet, and yet, there are indirect ways. We'll talk more about this in Week 7 (Evolution), but reflect on this: There is no doubt that mirror-capacity is a product of evolution: It conferred huge benefits for (altricial) bird and mammal parents to be able to monitor and provide for the survival needs of their offspring: Some of that is learned, but a lot of it is inborn (which means "learned" by natural selection), because perceptiveness of the cues to the needs of helpless offspring is as crucial for reproductive success as reproduction itself.
So, although evolution can no more mind-read than parents can, the relationship between offspring behaviour and parental perception is a mirror relation, both sides of which natural selection can modify and adapt.
Reflect on that. And also how it can apply to kinship, social behaviour and even predation.
Maya, good questions, but don't conflate (1) the correlation between behaviour and brain activity (where/when) in trying to reverse-engineer how the brain does what the organism does with (2) the fact that correlation does not explain causation. These are not the same problem. (And both these problems are equally present for the Easy Problem and the Hard Problem.)
Rena, mirror-capacities do not solve the Hard Problem of explaining how and why (sentient) organisms feel. Remind yourself that, with the same tools, Darwinian evolution can produce parental mind-reading in birds and mammals, and adaptive camouflage in insect colouring, or even in (presumably insentient) plant adaptations to climate change.
And even in Week 5, symbol-grounding (T3) is still part of the Turing programme. Turing's reverse-engineering has no way to handle feeling, which is unobservable; only the behavioral (and neural, T4) correlates of feeling are observable.
While reading this article I found myself wondering would a machine ever be able to fully capture the type of brain activity described here. What makes humans so unique in our ability to do what we do is in the complexity of our neural system like mirror neurons. As the article explains, mirror neurons are a distinctive class of neurons that discharge both when someone executes a motor act and when they observe another individual performing a same or similar act. Through studies they have been found to also help one understand the intentions of another’s actions, what others are feeling, and could have even lead to the evolution of speech. This feels incredibly similar to the predictable nature of ML and AI. Take for example a T3 system, it is able to “ground” it’s symbols in what it is experiencing through touch, vision, etc. (I use “ground” very loosely because it is really just taking in symbols adjusting different weights based on the symbols it takes in) and then use this new information to better predict how others will act in their environments and in turn how itself should output symbols that allow it to interact better with the environment. However, I keep finding myself more convinced that the T3 machine, while it is simulating a similar understanding as us, does not actually carry understanding of what the symbols it is processing from its environment actually mean. So while a T3 system would see someone picking apples and use symbol computations to make a prediction that they will eat the apple (just another symbol outputted) a person would have their mirror neurons activated and would actually understand not just what it means to do the action but also the links between the observed action and other actions related to it.
ReplyDeleteSierra, remember that Turing agreed from the outset (in the passage referring to “solipsism” [which was the wrong word for the O-MP]) that his method could only reverse engineer what thinkers can do — not what (or whether) they can feel>/b>.
DeleteMirror neurons seem like our body and minds attempt at solving (or embarking on) the other minds problem. This may be a bit of a loose link with the other minds problem, but I think at its essence we are trying to ask “how can we understand what others feel” and MN networks are our brains' manner of answering that question, albeit with lots of error and bias. The existence of MNs shows that understanding one another fundamentally lies in a process of embodiment of another's actions/emotion.
ReplyDeleteBecause we clearly need to understand one another, but cannot peek into or experience others' minds', our bodies' way around the other mind's problem is to observe action in others’, play it out within ourselves, and make an educated guess on what to anticipate from others by what that action might evoke in ourselves. The Bonini et al article describes this as remapping other-related information onto primarily self-related brain structures. Interestingly, our internal representation of the action does not need to be physically manifested for it to evoke the associated feelings. It would be cumbersome to have to mimic action going on around us to understand it.
I like the way you describe mirror neurons as the brain’s workaround for the other minds problem. Since we can’t directly access what someone else is thinking or feeling, it makes sense that our system “runs” their actions and emotions on our own neural hardware, then uses that internal replay to guess at their state of mind. To me, that really underlines how understanding is rooted in embodiment rather than in detached symbol manipulation. It lines up with Harnad’s point that computation by itself isn’t enough; something like the mirror system adds the grounding that purely formal processes lack.
DeleteWhat struck me too was your note that we don’t have to physically act out what we observe for the simulation to generate meaning. Embodiment here doesn’t mean literal mimicry, but rather the brain’s ability to tie what we see to potential actions and emotions in ourselves. It makes me wonder if building mirror-like mechanisms into artificial systems could help them move closer to real understanding, instead of just processing symbols. Of course, like you said, even our own system is imperfect, full of shortcuts, biases, and errors which just shows that cognition isn’t clean or mechanical but messy and human.
I had also made the same connection between the “Other Minds Problem” and the function of mirror neurons! Earlier, Dr. Harnad had mentioned in passing that, in spite of René Descartes’ argument on the uncertainty of everything that is outside of the mind, we can assume that we aren’t the only thinking person in a universe of other organisms that mimic our responses to simulate thought but cannot think. Although I agreed with the sentiment instinctually, I never understood why we can dismiss the possibility that no one else feels/thinks besides ourselves without empirical evidence. However, with the presence of mirror neurons that code intentionality behind action, I agree with your statement on the fact that may be the brain’s channel to “mind reading”. I would even raise the strength of the correlation and argue that their existence may be proof that other people think just as we do. Why else would we, as a species, have developed a region in the brain dedicated to understanding what people are thinking (sort of)?
DeleteI think the controversial part of neuroscience as it relates to cognitive science is that most of the data it provides is just data. There is plenty of evidence for correlation between certain brain activity and certain behavioural outputs, but it provides no answers to why this indicates something like understanding, intention or some other weasel-word. In a way, the brain as explained by neuroscience functions like a Chinese Room.
ReplyDeleteIn the example of mirror neurons, they are relating the perception of an action being done to the performance of doing the action themselves. But they don’t explain anything about why one is doing the action, whether there is understanding of or intention for what is being done, or even whether one is thinking at all (of doing the action or something unrelated).
With the heart, the story is simple because it has a primary job, to pump blood. The brain, on the other hand, doesn’t have just one or two main jobs. It ‘pumps’ everything we do from thinking, to planning, and speaking. Does that mean that the physical neurons themselves don’t really matter for understanding thought? If yes, then cognition is more like software (and you don’t necessarily need to know the details of the hardware to understand the program it’s running). However, unlike a laptop or smartphone, the brain isn’t a technological machine, it's an evolved tissue. The biology of neurons shapes how we perceive and act, and mirror neurons prove that action isn't just visual but it's wired directly into our motor system which we would only know from attempting to understand neurons and the brain itself. The debate is really surrounding how we each choose to understand ourselves, for example, do we want to explain thought in terms of abstract ‘computations’ or do we try to use neurobiology to explain why those computations are even possible.
ReplyDeleteLauren, I think you're right with your point about cognition as “software” is very interesting, but the mirror neuron research advocates for keeping hardware and software distinct. If premotor cortex neurons are activated when we grasp and when we see another person grasp, then biology is already doing work thats interpretive; there is no passive wiring occurring. Thus, mirror neurons show that the actual substrate not only "hosts" cognition, but it also organizes the very action categories and intention categories on which cognition operates.
DeleteThis research on mirror neurons links to Searle’s paper, adding evidence that computation is insufficient to represent cognition. It reminded me of the “Robot Reply”, which admitted how interaction with the environment is a vital aspect of our understanding of it. The evidence mirror neurons supports this, showing that activation of the motor system may be important to understand others’ actions, and that a mere visual perception, without involvement of the motor system would only provide a description of the visible aspects of the movements, which is insufficient to understand them.
ReplyDeleteInteracting with the world and experiencing for ourselves seems to be a key factor in much of human cognition, something that T2 cannot do.
Mirror-neuron-like mechanisms, which link the observation of an action to cortical motor representations, could be one of the components needed to create a T3 robot. It reminded me of the dynamic implementation of computation discussed in the 3b reading. Mirror neurons aid with environment interaction, as well as adaptation to others’ behaviour by activating our own motor cortex. This could in itself help with “grounding” the T3 robot in a sensorimotor interaction with its interlocutor, going beyond the “pen-pal” T2 robot. Mirror neurons might also play a role in the ability to learn of the T3 robot, by allowing it to analyze the behavior of the person in front of him and replicating it, therefore participating in the dynamic implementation of the computing sequence needed to accomplish a task.
ReplyDeleteThis is highly hypothetical, but I wonder if one way to test a T3's ability to appear undistinguishable from humans could be tested through a mirror therapy or experiment. Amputees can get relief from phantom pain by observing a reflection of their functioning limb on a mirror and moving it the way they would their aching missing limb. They know their limb is no longer there, yet it works. Non-amputees asked to put their hands on a table, with one behind a divider (out of sight) which is replaced by a rubber hand can be tricked into "feeling" the rubber hand as their own. To do so, the same touches are done to the hidden real hand, and the rubber hand the subject observes. They will recoil in fear reflexively when a hammer goes down on the rubber hand.
DeleteWould T3, whose actions are indistinguishable from ours, be fooled temporarily too if "amputated"?
“…visual perception, without involvement of the motor system would only provide a description of the visible aspects of the movements. It would not give, however, information on the intrinsic components of the observed action”
ReplyDeleteI actually find this line very interesting when it relates back to what we mentioned about the hard problem and how feelings can never be properly reverse-engineered. This provides a basis of why sensory-motor grounding is so important (T3) but, what I realized from this is that this connection simply did not exist it would become a much more difficult task for us grasp for an understanding of others’ intentions through images we see/our perception. This parallels our inability to truly know what it is to feel something, it could easily be like this case, where that path now between what we see or even the motor cortex and another part of our brain that might yield “feelings” (forgive me I’m not a neuroscience major) might not properly exist to give us the ability to fully understand feelings.
This reading made me think back on our definition of cognition and how we distinguish cognitive vs. vegetative function. I was wondering if the action of mirror neurons could act as a transition between the non-cognitive action of the brain, which happens whether or not we are thinking about it, and the cognitive action of the brain that thinks and manipulates information. As the mirror neurons take the raw perception from our visual system and map it onto a motor understanding, without any intentional choice to do so, they prepare the information to be thought about by cognitive systems.
ReplyDelete
ReplyDelete“A series of hypotheses such as action understanding, imitation, intention understanding, and empathy have been put forward to explain the functional role of the mirror neurons. In addition to these, it has also been suggested that the mirror neuron system represents the basic neural mechanism from which language evolved.”
If the mirror neuron system represents the basic neural mechanism from which language evolved, rooted in imitation, intention understanding, and empathy, then it raises a problem for the T2 system.
A T2 system, which operates only through symbols and rules, has no access to sensory information, motor resonance, or the empathic grounding that humans rely on to develop and use language. This tension invites a critical question. If human language is linked to embodied simulation and social resonance, how can a purely syntactic system like T2 legitimately be said to “have” language? Does this mean the Turing framework captures only the outward form of language, while missing the embodied and empathic foundations that give human communication its depth? And if so, how does this limitation feed into the hard problem, and the relationship between language, computation and cognition?
This article about mirror neurons made me think of the concepts of awareness versus self-awareness and wonder how mirror neurons might be involved. We are quite sure that many different species of animals with nervous systems are aware / conscious / sentient (whatever you may call it), but of those many species, only a select few of them seem to be able to consistently recognize their reflection as themselves in a mirror. This ability is sometimes argued to be proof of self-awareness (i.e. the ability to become the object of one’s own attention), an important piece of human cognition. If mirror neurons play an important role in action and intention understanding, then might they also be a key component of the rare ability for self-awareness in animals when observing the actions of their own reflections?
ReplyDeleteMirror neurons surely play a part in explaining cognition. In his book From Bacteria to Bach and Back, Dennett argues that one reason why we have minds or consciousness is due to our social nature. Indeed, there could have been a selective pressure in humanity’s evolution for humans who better understand other humans. After all, if you know who you can trust, who is lying, understand hidden intentions, you have a better chance of survival and passing genetic materials. Although do not fully understand MN, it would be unsurprising to see MN networks activating when we see someone trying to deceive us. If it is really the case, understanding MN will be crucial to understanding how neural activity leads to cognition.
ReplyDeleteThe article introduces a class of specialized neurons known as mirror neurons, which become active not only when an individual performs a specific motor act but also when observing someone else perform the same act. This extends into the domain of empathy, defined as “the capacity of feeling the same emotions that others feel”. Indeed, fMRI studies have demonstrated neural resonance for motor intentions, disgust, and physical pain. For example, when you see someone stub their toe, the same neural circuits that would activate if you stubbed your own toe light up. In relation to the hard problem of consciousness (i.e., how and why we feel the way we do), mirror neurons may bring us closer to understanding how we can simulate or resonate with others’ states. However, this still leaves us wondering about the “why” – that is, why neural firings translate into specific felt experiences such as pain or sorrow, or into subjective feeling at all.
ReplyDelete“A mere visual perception, without involvement of the motor system would only provide a description of the visible aspects of the movements. It would not give, however, information on the intrinsic components of the observed action, on what it means to do the action, and on the links between the observed action and other actions related to it.”
ReplyDeleteI think this passage is saying that vision on its own only shows the outside of an action, and that mirror neurons are what really let us grasp its meaning. The argument is that true understanding comes from mapping what we see onto our own motor system (so that observing feels closer to doing). I find that pretty convincing because it highlights a deeper link between watching, doing, and understanding. However, I do think the passage underestimates how much vision can tell us by itself. From my own perspective, I feel like we can (a lot of the time) figure out what people are doing just by looking. Context, facial expressions, and the way someone moves can tell me a lot, even if I’m not mirroring them in my own body.
I find it interesting that mirror neurons can address the problem of direct comprehension. The authors describe this as messages that are "understood by an observer without any cognitive mediation." This can be because the motor representation in the parieto-frontal mirror system is the same for the sender and recipient. As such, mirror neurons allow for direct understanding between the message sender and receiver. The authors note that this is done without the need for “arbitrary” symbols. Does this imply that empathy is not computation, that it’s more inherent?
ReplyDeleteI read the “no arbitrary symbols” part as meaning mirror neurons let us understand actions directly, without translating them into learned codes–if I see someone grasp an apple, my own motor system activates as if I were grasping it.
DeleteThat reminds me of Jenna’s point about vegetative versus cognitive capacities. Mirror neurons seem to work at that more basic level—they give us an instant, built-in response that helps us directly pick up on what someone else is doing. But like others (and Prof. Harnad) have pointed out, this doesn’t get us all the way to empathy. We can never feel what someone else is feeling with certainty--our own experiences are the only ones we have direct access to.
And I think this connects back to the Turing-style thought experiments. Even if we could “build in” mirror neurons in a T2 or T3 system, what we’d really have is just a symbol for a mirror neuron, not the direct neural link between sender and receiver that Rizzolatti & Destro describe. So mirror neurons might help explain a foundational layer of communication and empathy, but they don’t collapse the bigger problem of how meaning and genuine understanding arise.
The Turing Test asks whether a machine can imitate human conversational behaviour so convincingly that an interrogator cannot tell it apart from a person, but passing this test may require more than surface-level pattern matching. Mirror-neuron research shows that in humans, understanding another’s action involves a direct internal coupling between perception and execution. The observer’s brain partially simulates what it sees. Communication, then, is not only verbal but deeply embodied, involving gestures, tone, and posture. If a machine merely processes symbols without this kind of embodied mirroring, it may produce credible words, yet lack the underlying empathic resonance and contextual attunement that human interaction relies on.
ReplyDeleteThis raises a key question, to truly pass an expanded Turing Test that includes social and nonverbal cues, would a machine need its own “mirror system,” and how could we tell whether such mirroring reflects genuine understanding rather than sophisticated imitation?
Ayla, I think you make a great point, and the question you raised is really important. It reminds me of a debate we had in one of my psychology classes about whether AI could ever replace therapists in the future. I think it comes back to your question: if AI can mirror human behaviour, like a therapist does, does it genuinely understand what the patient is saying, or is it simply imitating responses stored in the machine? I don’t think machines can show genuine understanding because they can’t truly perceive and interpret nonverbal cues, which humans process through mirror neurons that enable empathy. Since AI lacks this kind of sensorimotor grounding, it might be able to engage in therapy linguistically and produce responses that sound human, but this doesn’t equate to genuine understanding.
DeleteOne thing that stood out to me is that mirror neurons fire not only when performing an action but also when observing it. Some researchers argue this points to a deeper role in action understanding rather than just motor resonance. But it makes me wonder, if mirror neurons are tied to understanding the goal of an action, how do we separate whats just a lower level mimicry from genuine goal inference? Could mirror activity alone explain intention reading, or does it need to be combined with higher order cognitive systems?
ReplyDeleteI see what you are saying - that is that mirror neurons fire both when we act and when we watch and that it could just be advanced mimicry vs real goal understanding. In the reading they push back on that by saying that while mirroring looks like mimicry its the pattern of neural activation that suggests something deeper (our brains encoding the goal of the action and not just the movement itself). Simple mimicry like emotional contagion (a baby crying when another cries or automatic yawning) is defined as "simple forms of behavioural sharing" which the authors state isn't true empathy or goal grasp. (pg1 Corradini and Antonietti) Mirroring alone (type 1 process -fast, automatic) isn't enough in goal inference because the true comprehension requires the involvement of higher-order cognitive systems defined as type 2. This higher level capacity and complexity is what the authors call reenactive empathy or mentalising, which necessitates slow, elaborate inferences and abstract reasoning.
Delete
ReplyDeleteTo me, there is a lot of weight placed on goal-directedness. The Scholarpedia article emphasizes that mirror neurons “map the pictorial description of actions… onto their motor counterpart,” but only reliably when there’s a clear object or goal involved. That made me wonder: maybe the mirror system isn’t just “copying” others’ actions, but actually compressing a wide range of movements into a smaller set of goal-codes. If this is true, it would explain why pantomimes or ambiguous gestures don’t trigger strong, stable mirroring, maybe they don’t compress neatly.
This “compression” view shifts the function away from empathy or imitation toward something more computational: reducing messy visual input into a compact, goal-based representation that can guide prediction and learning. It also raises testable questions: does mirror activity have lower dimensionality than the visual input it receives? And can training with new tools expand the system’s codebook? If so, mirror neurons aren’t about “duplication” at all, but about goal-selective abstraction.
Mirror neurons (neurons that fire both when you are doing an action or seeing someone else do the same action) offer a lot of potential for learning the inner mechanisms on how we get to learn the process of doing. Specifically, I found the support for the gestural theory of language evolution ( language evolved from doing gestures and eventually these gestures were associated with sounds) very interesting. In addition, we know that infants are able to learn language through observation , so having a cluster of neurons that seem to correlate with both observation and action is a start in understanding the process of language learning. However, study of the brain is not a replacement for reverse engineering mental processes to study cognition. As was mentioned above. the brain already does so much at all times that we can't reliably explain everything nor explain the why and how. Also, the neuroplasticity of the brain (how brain areas can take over when other areas are damaged) makes it so that these specific brain areas might not be as crucial to these processes as we would like to believe. Mirror capacities could also be something that we got over time which then would require us to figure out how we acquired it (back to square 1).
ReplyDeleteOne question that I found myself asking while reading the paper on mirror neurons was that does artificial intelligence stimulate mirror neurons in the same way that other humans do? Bratan et al found that the appropriate mirror neurons did not fire when listening to voices created by AI. This finding can potentially inform a subset of the T3, whereby we also measure whether the robot stimulates the mirror neurons of the person it is interacting with.
ReplyDeleteHere is the paper that I mentioned: https://www.researchgate.net/publication/394142521_Mirror_Neurons_cannot_be_Fooled_by_Artificial_Voices_-_a_study_with_Implications_for_Education_using_Magnetic_Resonance_Imaging_MRI_and_Convolutional_Neural_Network_CNN
I don't think mirror neurons are irrelevant, but I do get what Harnad means when he says they don't explain cognition. They provide us with insight into the relationship between perception and action that goes beyond simple correlations. They at least reduce what must be taken into consideration in a complete explanation. When I move and when I watch someone else move, the same circuits fire. This is not insignificant; rather, it is a structural hint as to how prediction might operate. These neurons don’t just activate when something happens, they seem to prepare for what’s coming next, as if they already have a sense of the goal behind the action. That doesn’t feel like “understanding” in the full sense, but it shows that the brain’s engagement with the world is already predictive and embodied. This makes me wonder what an expanded Turing Test would look for—one that measures not just linguistic fluency but predictive ability. Could a GPT ever mirror the goals behind behaviour, rather than just the behaviour itself? Would that count as understanding, or only as a more refined form of imitation?
ReplyDeleteWhilst reading this article, the Fogassi et al. (2005) study caught my interest as it found that in monkeys, there are IPL neurons that fire selectively for certain actions like grasping with the intent to eat and grasping with the intent to place. This suggest that the brain can infer the intentions of others and even predict their next action. Therefore, if we consider this, do you think that the ability of the brain to interpret goal-directed action through neural activity could’ve led to the basis of gesture-based languages like ASL and BSL?
ReplyDelete