cover encarni marcos

Encarni Marcos on prefrontal cortex and decision making

  • cover play_arrow

    PLAY EPISODE


cover encarni marcos
Season 2019
Season 2019
Description arrow_drop_down

Description

Why do some prefrontal neurons hold steady while others rapidly switch what they represent? Neuroscientist Encarni Marcos reveals that the prefrontal cortex operates through a continuum of neural stability and flexibility , where heterogeneous populations simultaneously maintain goals in memory and dynamically transform them into actions. Subscribe for more from the Convergent Science Network podcast series. Encarni Marcos joins Paul Verschure and Tony Prescott to discuss her research on how prefrontal cortex supports goal-directed behavior. Recording from dorsolateral prefrontal cortex in monkeys performing discrimination tasks, she finds that neurons do not simply encode one feature of a task. Instead, individual neurons represent multiple features, goals, cues, actions, often overlapping in time, with some neurons switching their representational allegiance as a decision unfolds while others remain locked to a single variable throughout the trial. The conversation explores what this heterogeneity means for decision-making. Marcos describes a model built from competing pools of neurons with different excitability levels: stable populations maintain task-relevant information as a kind of ground truth, while flexible populations reshape network dynamics to drive the transition from goal representation to action selection. This architecture, validated against physiological data including burst-pause patterns, offers a mechanistic account of how the brain can simultaneously remember what it needs to do and figure out how to do it , without requiring separate memory and decision systems. Key topics include why averaging across neural populations obscures the real dynamics of prefrontal cortex, how error signals in prefrontal neurons defy standard dopaminergic prediction error models, the limitations of drift-diffusion models for explaining individual neural dynamics, and why neural variability may carry more information about cognitive processing than firing rates alone. Part of the Convergent Science Network podcast series from the BCBT Summer School.

Tagged as:

About the author call_made

CSN Podcasts

Both the triumphs of humanity and its most evil deeds have resulted from collaboration. In a time where humanity is required to aspire to the former and minimize the latter, the question arises of how collaboration arises and why it fails. Surprisingly, this phenomenon, so central to who we are, is not well understood. Hence, a collaborative effort is required to understand collaboration in its full biological, psychological, sociological, cultural, and economic complexity and to translate this understanding into operational impact. This series of podcasts is one step toward achieving these complementary goals. The Collaboration Podcast presents interviews with people who are central orchestrators of collaboration in various domains including business, government, science, art, health, sustainability, and the military. The discussions were conducted by Prof. Dr. Paul F.M.J. Verschure and members of the Program Advisory Committee of the Ernst Strungmann Forum on Collaboration (https://www.esforum.de/forums/ESF32_Collaboration.html) during 2021 and had the goal to sketch a map of opportunities, challenges, and obstacles in human collaboration. The forum took place in May 2022, and now we would like to share this series of interviews with a broader audience. The full report of the Forum will be published in 2023 by MIT Press. The podcast was produced by the Convergent Science Network (https://www.convergentsciencenetwork.org/). Context: The stability of social systems depends critically on realizing sustainable methods of “collaboration,” yet how and by which means collaboration is achieved is not clearly understood; neither are the conditions or processes that lead to its breakdown or failure. Collaboration can be understood as cooperation between agents toward mutually constructed goals. Part of the reason for our lack of understanding is that the phenomenon of collaboration is, by nature, a highly multidisciplinary problem, and effective research into its complexities has been difficult to achieve across the broad range of scientific and technical disciplines involved. The need for a fundamental understanding of collaboration, however, has become increasingly important. Not only does humankind demand answers as it attempts to address critical challenges at multiple scales (e.g., climate change, migration, enhanced automation, social and economic inequality), but ever-increasing technological and economic means of interconnecting people and societies are disrupting long-established, familiar patterns of how we interact. Radical technological changes that are ongoing have the potential to reshape collaboration in ways that are currently hard to predict or influence (e.g., by altering configurations in interaction, information creation, and modes of communication). On one hand, such changes could disrupt hitherto stable forms of collaboration by affecting critical communication channels and traditional roles, as can be observed in the rapidly changing patterns in governance, commerce, and social interaction. Conversely, technology could lead to the emergence of novel, successful forms of collaboration that deviate from traditional “hierarchical” architectures. Evidence of this can be seen in areas as diverse as highly automated manufacturing plants, the open science movement, collaborative software repositories, user-centered services, and the sharing of economy-based modes of organization. Without a fundamental understanding of the mechanisms, processes, and boundary conditions of collaboration, it is not possible to evaluate or predict which of these possible scenarios are sustainable or even plausible. The Forum “How Collaboration Arises and Why it Fails” (May 8–13, 2022, Location: Frankfurt am Main, Germany) Chairs: Andreas Roepstorff and Paul Verschure Program Advisory Committee: Jenna Bednar, Julia R. Lupp, Bhavani R. Rao , Andreas Roepstorff, Ferdinand von Siemens, and Paul Verschure

More posts

Timestamp

  • fast_forward00:00:03 - This is the Convergent Science Network podcast. Leading researchers in the domain
  • fast_forward00:00:10 - of neuroscience, brain theory and technology are interviewed by Paul Vershoor and Tony Prescott.
  • fast_forward00:00:20 - So this is Paul Vershoor for the Convergent Science Network podcast and we're
  • fast_forward00:00:25 - here at the Barcelona Cooperation Brain and Technology Summer School of 2018,
  • fast_forward00:00:30 - and as one of our speakers today we have Ancordi Marquos Hello,
  • fast_forward00:00:36 - Ancordi so you were talking about two things like you emphasized very much this
  • fast_forward00:00:42 - whole idea of how prefrontal cortex represents the memory of goals and initiates actions right so,
  • fast_forward00:00:54 - So what do you see as sort of the key feature of these neurons that needs to be explained?
  • fast_forward00:01:00 - I think the most important result or the most important thing that we have seen
  • fast_forward00:01:05 - with data is that the neural network in the frontal cortex is really heterogeneous.
  • fast_forward00:01:11 - So you cannot explain anything just by looking at the average finding rate of
  • fast_forward00:01:16 - depopulation because you have nothing there.
  • fast_forward00:01:18 - You have really to go inside the individual dynamics to be able to see something
  • fast_forward00:01:24 - and to really understand what is going on there.
  • fast_forward00:01:27 - Right, so what are the key observations that you then start out from in this exploration?
  • fast_forward00:01:36 - So one of the key results is that we found that there are some neurons or some
  • fast_forward00:01:41 - group of neurons in the prefrontal cortex which really play an important role
  • fast_forward00:01:45 - in shaping the whole dynamics of the network.
  • fast_forward00:01:48 - And these neurons are so important because they are very susceptible to the
  • fast_forward00:01:52 - input so that they change state very easily.
  • fast_forward00:01:55 - And that, but consequently, they can shape the whole network activity and we
  • fast_forward00:02:01 - can be flexible about what we do or how we interact with the world.
  • fast_forward00:02:07 - While in the same time, we have neurons which are much more stable so that they
  • fast_forward00:02:10 - keep track of what we did and which is our internal state or all the information
  • fast_forward00:02:16 - that could be relevant for your way of doing it. Right.
  • fast_forward00:02:20 - But one of the first experiments that you discussed was actually an interesting one,
  • fast_forward00:02:26 - where you looked at how neurons responded to the feedback that the animal received
  • fast_forward00:02:32 - after it performed a task, correctly or incorrectly.
  • fast_forward00:02:37 - And you showed that if you measure from dorsolateral prefrontal cortex.
  • fast_forward00:02:42 - Neurons came actually two flavors, right?
  • fast_forward00:02:45 - That some showed an elevated response and others actually did not change the
  • fast_forward00:02:52 - response level after the feedback was received.
  • fast_forward00:02:55 - But what was curious there is that you showed that these neurons increased their response.
  • fast_forward00:03:02 - So I initiate a movement, I get my feedback, I make an error,
  • fast_forward00:03:07 - and the neurons, there were neurons that actually elevated their response when
  • fast_forward00:03:11 - an error was made, right?
  • fast_forward00:03:13 - As opposed to neurons that actually did not change at all in case the response was correct.
  • fast_forward00:03:20 - So this was a little bit the starting point to show that, okay,
  • fast_forward00:03:23 - there is sort of an internal memory system at work, but it elevates the response when I make a mistake.
  • fast_forward00:03:31 - So what's the significance of that?
  • fast_forward00:03:35 - Well, in the task in which we did the analysis, actually, we saw that it was
  • fast_forward00:03:42 - not really very much there because they won't use that information at all in
  • fast_forward00:03:46 - their subsequent behavior.
  • fast_forward00:03:49 - But we think that this might be very important in dynamic environments or in
  • fast_forward00:03:54 - situations in which you have to be flexible, actually, to adapt to changes.
  • fast_forward00:04:00 - So you need to know if you did well or not, because that could have an influence
  • fast_forward00:04:04 - in your subsequent behavior.
  • fast_forward00:04:06 - So we think that the different neurons are anyway monitoring this information,
  • fast_forward00:04:12 - in case there are some changes or in case this could be relevant for something else later on.
  • fast_forward00:04:18 - In this case, this elevated response in the firing rate to an error trap.
  • fast_forward00:04:26 - It might not be of immediate relevance
  • fast_forward00:04:28 - in that task because you're not controlling in any way error, right?
  • fast_forward00:04:35 - But could you still argue that maybe for other mnemonic functions this error
  • fast_forward00:04:40 - function might play a role in some other process that is not directly translated to task performance?
  • fast_forward00:04:47 - Or would you really say irrelevant? No, no. I say irrelevant for what we looked at.
  • fast_forward00:04:53 - I don't think it's irrelevant for everything.
  • fast_forward00:04:56 - So I think if the prefrontal cortex or the brain in general is monitoring and
  • fast_forward00:05:01 - monitoring this information or keeping track of that is because it might be useful at some point.
  • fast_forward00:05:08 - For this task in particular, we didn't find any correlation between incorrect
  • fast_forward00:05:14 - and the migration of behavior, but it could be relevant if in the same task
  • fast_forward00:05:20 - we change some, unexpectedly,
  • fast_forward00:05:22 - we change something of the task.
  • fast_forward00:05:25 - The more people use this information, the more adaptation.
  • fast_forward00:05:29 - So, it's not irrelevant in general, but it's, yeah, maybe the word was not well used.
  • fast_forward00:05:36 - It's not irrelevant for everything, but it's not relevant in the way the monk
  • fast_forward00:05:41 - is using that information in that task.
  • fast_forward00:05:43 - I think it doesn't mean that it's irrelevant.
  • fast_forward00:05:48 - So what's interesting as well is that usually there's an interpretation that
  • fast_forward00:05:53 - that feedback to the NNL in error is translated to some sort of modulatory signal
  • fast_forward00:05:59 - to prefrontal cortex, right?
  • fast_forward00:06:00 - Where you would say, well, I got an unexpected reward, dopamine goes up transiently,
  • fast_forward00:06:05 - and I should see some change in the response.
  • fast_forward00:06:07 - I have an unexpected error, and I should see a transient downregulation of dopamine,
  • fast_forward00:06:12 - which should also be translating to a transient response change in the neurons.
  • fast_forward00:06:18 - But in some sense, you see something very different now, because we see here
  • fast_forward00:06:23 - now that an error translates into an increased response, which would be associated with, let's say,
  • fast_forward00:06:31 - release of dopamine, but that doesn't make sense given this standard model.
  • fast_forward00:06:36 - The standard model says, no, the not expected error should be a reduction in
  • fast_forward00:06:40 - dopamine response, right?
  • fast_forward00:06:41 - So, from that perspective of how I've been thinking about how error regulates
  • fast_forward00:06:47 - prefrontal cortical responses,
  • fast_forward00:06:49 - is this a surprising outcome or you think it's really sort of fixed in the standard
  • fast_forward00:06:54 - interpretation of how this system should work?
  • fast_forward00:06:58 - For us, it was at first, it was surprising.
  • fast_forward00:07:02 - We were not expecting that the activity would increase for the employers,
  • fast_forward00:07:06 - but as you said, we would expect it to increase for the, if there is a double
  • fast_forward00:07:11 - increase, we were expecting it to increase for the reward delivery.
  • fast_forward00:07:14 - But we found actually the opposite.
  • fast_forward00:07:21 - We found it, we tried to find some kind of correlation actually between the
  • fast_forward00:07:26 - error and and the difficulty of the trial, to say, okay, do we have some correlation
  • fast_forward00:07:31 - where that's maybe for more difficult trials?
  • fast_forward00:07:35 - You don't have this activation in the neurons because you were just expected
  • fast_forward00:07:40 - to be wrong, so you just have nothing.
  • fast_forward00:07:42 - But for the easiest trials, for some reason that we cannot really explain yet,
  • fast_forward00:07:48 - you have this activation because in some places it's like, hey,
  • fast_forward00:07:52 - look, this was wrong, and I was not expecting that.
  • fast_forward00:07:54 - But we don't speculate yet because we cannot answer that in which are the mechanisms
  • fast_forward00:08:00 - causing this but actually we didn't find either this correlation with difficulty
  • fast_forward00:08:05 - so we could not explain that in that way so,
  • fast_forward00:08:09 - yeah it was a surprising result and,
  • fast_forward00:08:13 - from what we have now we cannot go more into the details of that Is that fair
  • fast_forward00:08:19 - to say that the question is a standard model or do you Do you believe it will
  • fast_forward00:08:24 - be incorporated in the standard model if you just have a bit more time to look at the details of this?
  • fast_forward00:08:31 - I think this should be incorporated into the standard model.
  • fast_forward00:08:34 - I don't think it's against the standard model.
  • fast_forward00:08:37 - I think this is some feature that should be taken into account. Okay.
  • fast_forward00:08:41 - So now we know that we have responses at different timescales.
  • fast_forward00:08:47 - We have very systematic responses even after the action has been emitted,
  • fast_forward00:08:52 - after we've received our feedback, in this case, that relates to this error, the error component.
  • fast_forward00:08:56 - But then sometimes you were delving much more in detail in that,
  • fast_forward00:08:59 - And you started to look at, okay, how does the response of these prefrontal cortical neurons,
  • fast_forward00:09:06 - if a simple discrimination test, whether to say whether something is closer or further away,
  • fast_forward00:09:11 - or whether a time interval is shorter or longer, with the simple discrimination
  • fast_forward00:09:16 - test, you started to see that these memory-based responses of these neurons
  • fast_forward00:09:21 - were actually not univariate.
  • fast_forward00:09:25 - It was not that they were just reflecting one property of the task,
  • fast_forward00:09:28 - right? They could also switch, if you want, their response properties.
  • fast_forward00:09:33 - So what's really the implication of that?
  • fast_forward00:09:38 - So how dramatic is that, this switching of the response properties?
  • fast_forward00:09:43 - Do you mean from calling something memory to calling something else afterwards?
  • fast_forward00:09:48 - Yeah. How dramatic? Yeah, so this was the experiment that you published in Psychedelic
  • fast_forward00:09:55 - Reports in 2016, if I'm correct.
  • fast_forward00:09:58 - Yes. So where you show these persistent activities in prefrontal cortex are not committed.
  • fast_forward00:10:04 - If you look at the single cell, it's not that the single cell has high fidelity
  • fast_forward00:10:08 - to one property of the task.
  • fast_forward00:10:10 - It might initially maybe code for what the goal is of the task,
  • fast_forward00:10:14 - but at some point it just switches its allegiance and starts to code something else.
  • fast_forward00:10:18 - It starts to go in the action you want to transmit.
  • fast_forward00:10:22 - So how fundamental is that property, this ability to switch?
  • fast_forward00:10:25 - Do you see that in many of these neurons, or is it something that is sort of very, very rare?
  • fast_forward00:10:31 - No, actually, this is quite common in different types of neurons.
  • fast_forward00:10:35 - So they are not just tuned to one of the properties of the task, but normally they code.
  • fast_forward00:10:39 - The normal thing is that they represent more than one feature of the task.
  • fast_forward00:10:45 - So it's quite common to find that in the prefrontal code.
  • fast_forward00:10:48 - But they code one feature in sequence, right? So at one period of time,
  • fast_forward00:10:52 - they code one thing, and at another period of time, they code the other thing? No.
  • fast_forward00:10:55 - Or are they more multiplexing? No, it could be overlapped, too.
  • fast_forward00:10:59 - So you could have, for instance, in the result that I presented about the transformation
  • fast_forward00:11:04 - from goal in memory to the action, there you see that some neurons score the
  • fast_forward00:11:09 - goal at the same time as the action. So they overlap.
  • fast_forward00:11:14 - They have a bit of delay, but during some period, both things are going by the
  • fast_forward00:11:19 - same level. If they switch their representational connotation,
  • fast_forward00:11:27 - do you see that also as reflecting the progression of the decision-making?
  • fast_forward00:11:32 - Like, initially I have to make up my mind of what are my options,
  • fast_forward00:11:36 - what's the relevant evidence, what's memory telling me?
  • fast_forward00:11:39 - So at this point in time, you want to dedicate more representation or resource
  • fast_forward00:11:43 - to what's the goal I'm pursuing.
  • fast_forward00:11:46 - But maybe at some point when you have pruned down your options and you say,
  • fast_forward00:11:50 - okay, to achieve the goal, this is now my action, or at least a subset of actions
  • fast_forward00:11:54 - I should consider, maybe that in the progression of the decision process,
  • fast_forward00:11:59 - I might want to shift my representational resource.
  • fast_forward00:12:03 - Is that also how you see that? Yeah, I find this to be correctly,
  • fast_forward00:12:07 - yeah, we found it like a sequence of states.
  • fast_forward00:12:10 - So first you have something in memory and because of the, because you are iterating
  • fast_forward00:12:15 - more and more information now, I should see which action to perform,
  • fast_forward00:12:19 - so neurons are switching the tuning or the faring towards that feature because
  • fast_forward00:12:25 - it's the one which is most relevant at that point.
  • fast_forward00:12:27 - So it's more or less a sequentiality of states.
  • fast_forward00:12:30 - Actually, there are some works that are only with Hilleman of Mode that what
  • fast_forward00:12:37 - they saw is actually that,
  • fast_forward00:12:38 - that we have sequences of states through which the neural population in the
  • fast_forward00:12:42 - prefrontal cortex and also in other cortical areas Yes, also through sequences of states,
  • fast_forward00:12:48 - so that they are adapting to the new events or new features coming in.
  • fast_forward00:12:53 - But does it reflect only that new information comes in, or does it also reflect
  • fast_forward00:12:58 - the progression of the decision-making process?
  • fast_forward00:13:01 - Both. Also the progression of the decision-making process.
  • fast_forward00:13:04 - You can have changes in the states as the decision is being made,
  • fast_forward00:13:10 - and also the causes are intervening on some new information.
  • fast_forward00:13:13 - So both cases are different.
  • fast_forward00:13:15 - Right. So, most of your neurons are in dorsolateral prefrontal cortex, right?
  • fast_forward00:13:22 - Do you think that this is a generic feature also if you go to medial frontal cortex?
  • fast_forward00:13:27 - Or would you feel it's really a more specialized property of dorsolateral prefrontal cortex?
  • fast_forward00:13:34 - Well, this I cannot respond with a clear answer because we didn't know about that.
  • fast_forward00:13:44 - But my guess would be that you might find it also in a way in the asymptotic
  • fast_forward00:13:49 - or the saloprilage frontal cortex.
  • fast_forward00:13:52 - I think it's more likely that they call it a ticking voice in the brain of some changes. Right.
  • fast_forward00:13:58 - But so, then in this first experiment, you started to focus more also on how
  • fast_forward00:14:03 - populations of neurons are actually then tuning themselves to the task, right?
  • fast_forward00:14:08 - And from that came also this idea that some populations are showing this progression
  • fast_forward00:14:15 - in sort of a task-dependent way, and other populations do not show this progression.
  • fast_forward00:14:23 - When you show these groups of neurons that you were measuring from,
  • fast_forward00:14:29 - then you have the pre-go and post-go goal neurons.
  • fast_forward00:14:35 - Neurons, but some of those are actually switching their representational state, and others do not.
  • fast_forward00:14:42 - So, how should I think about that? How should I interpret that? What does that mean?
  • fast_forward00:14:47 - Also, in the face of what we just discussed, that there is this sort of progression of decision-making,
  • fast_forward00:14:52 - the resource is reallocated in some sense, but should I look at this result
  • fast_forward00:14:58 - as telling us that, okay, I have a bunch of neurons in the prefrontal cortex?
  • fast_forward00:15:04 - In the end, they're multiplexing because across all these neurons you will find
  • fast_forward00:15:08 - that they represent all possible aspects of the task and all possible combinations of these aspects.
  • fast_forward00:15:15 - This is roughly what it means, right?
  • fast_forward00:15:17 - Some of these will then again switch what they represent, and others will not
  • fast_forward00:15:24 - switch what they represent, right?
  • fast_forward00:15:25 - So there's a fixed representational substrate, and others will more dynamically
  • fast_forward00:15:29 - allocate themselves to that.
  • fast_forward00:15:32 - So how should How should we interpret that? That there are these neurons that
  • fast_forward00:15:34 - are stupid, they're dynamic, and others are smarter and they're dynamic?
  • fast_forward00:15:38 - Or how do you interpret that? How we interpret it is, we have two main key features
  • fast_forward00:15:46 - in the prefrontal cortex.
  • fast_forward00:15:47 - One are composed by neurons, which are very stable in different ways,
  • fast_forward00:15:51 - so they are keeping track of what is going on, of the information which is important
  • fast_forward00:15:57 - for something, for the decision, for attention, for whatever.
  • fast_forward00:16:00 - There, so they are very stable and robust.
  • fast_forward00:16:02 - And then we have a second group of neurons, which are flexible because in the
  • fast_forward00:16:07 - end, you cannot just stick to something and that's it,
  • fast_forward00:16:10 - you need to be flexible to be able to perform an action or to list or to get
  • fast_forward00:16:16 - more information or whatever.
  • fast_forward00:16:18 - So these two dynamics are very important in the brain because one is telling
  • fast_forward00:16:22 - us, okay, this is the information we have.
  • fast_forward00:16:25 - And the other is saying, okay, but now we need to move from that information. We need to do something.
  • fast_forward00:16:30 - So the other ones, the ones which are shaping the whole nerve or any different
  • fast_forward00:16:34 - dark cortex to say, okay, now is the time to perform an action. Right.
  • fast_forward00:16:38 - But wouldn't it be fair then to say that there is one set of neurons,
  • fast_forward00:16:43 - let's say 50% of the population, gives you like a ground truth and says,
  • fast_forward00:16:47 - this is what is out there in the world.
  • fast_forward00:16:50 - This is what my needs are and my goals.
  • fast_forward00:16:53 - And I'm not going to, this is the blackboard in which I'm going to operate now. happen.
  • fast_forward00:16:57 - And they have another representational substrate or another set of neurons that
  • fast_forward00:17:00 - then say, okay, now if I want to come to a solution of this,
  • fast_forward00:17:04 - the other things I should be doing.
  • fast_forward00:17:06 - So it sort of moves away from that ground truth and now starts to sort of prune,
  • fast_forward00:17:10 - you know, okay, but this isn't the specific cue that matters.
  • fast_forward00:17:13 - And this isn't the specific thing I should be doing and I should ignore the
  • fast_forward00:17:16 - other stuff. Is this roughly how you would think about that?
  • fast_forward00:17:19 - Not exactly. Because we are talking always about two loop of neurons,
  • fast_forward00:17:22 - but these are two extremes.
  • fast_forward00:17:24 - We believe that in the prefrontal cortex we have a continuum of neurons.
  • fast_forward00:17:28 - So that means that it's not that we have only a loop of neurons that are very
  • fast_forward00:17:33 - stable and then one which is very flexible.
  • fast_forward00:17:36 - But that we have also something in the middle that goes from stability to flexibility.
  • fast_forward00:17:41 - So, when I talk about two groups, I'm always talking about the same cases,
  • fast_forward00:17:46 - but we believe that we have neurons doing things also in the middle range.
  • fast_forward00:17:51 - Okay, that is true, because indeed, in your data, you filtered quite a bit,
  • fast_forward00:17:54 - right? In the end, we only looked at the no-switchers or the switchers.
  • fast_forward00:17:58 - Exactly, because these were the strengths of the model. All right,
  • fast_forward00:18:01 - so it's interesting, because it means in all cases you represent all possible
  • fast_forward00:18:06 - aspects of the task and all its combinations,
  • fast_forward00:18:08 - but then also you represent all possible transformations of that over time.
  • fast_forward00:18:14 - Is that what you're saying?
  • fast_forward00:18:15 - So you're expanding the whole search space, actually. Yeah, yeah.
  • fast_forward00:18:21 - So then… For instance, with our model, we actually fitted the data that we were
  • fast_forward00:18:28 - looking at, but remember that we were always looking at 1,000 neurons,
  • fast_forward00:18:32 - and we are selecting a class 2 population of them because they are doing what we are looking at.
  • fast_forward00:18:37 - But of course, we have many different groups of neurons, and many neurons doing
  • fast_forward00:18:41 - all of this stuff. Right.
  • fast_forward00:18:43 - Exactly. No, this is, I understand. Okay. So it is, so we have a hyper,
  • fast_forward00:18:48 - high dimensional representation of your task and its changes, right?
  • fast_forward00:18:53 - So that you would believe there's a bit of a problem.
  • fast_forward00:18:56 - How do I now select among all these options?
  • fast_forward00:18:58 - How do you see the selection taking place?
  • fast_forward00:19:02 - And this also brings us a little bit to the model that you built of this, right?
  • fast_forward00:19:06 - So now I have this high dimensional space with all sorts of combinations of features of the task.
  • fast_forward00:19:15 - One is sort of constant, and the other is dynamically now, if you want,
  • fast_forward00:19:22 - trying options and weighing them, right?
  • fast_forward00:19:26 - So how do you see that process play out?
  • fast_forward00:19:35 - How should I imagine that this actually can lead to even a decision?
  • fast_forward00:19:40 - That it can converge to a decision on an action I should perform? All these neurons.
  • fast_forward00:19:45 - Yeah. This whole process, this whole dynamo process that you mentioned, Probe.
  • fast_forward00:19:49 - Well, they are, I mean, I could see that as, I mean, do you imagine that the
  • fast_forward00:19:55 - neurons which are stable and not falling apart are hitting back on the information
  • fast_forward00:19:58 - which might be important at that point?
  • fast_forward00:20:00 - And you see the other group of the flexible group getting more information from
  • fast_forward00:20:05 - outside, but also from our internal motivation, attention, and everything.
  • fast_forward00:20:09 - And you see this is a couple of connected areas and neurons inside the frontal corpus.
  • fast_forward00:20:18 - As I show in the presentation, just by connecting it with some excitation and emission,
  • fast_forward00:20:25 - you would get this selection at the end, because with the flexibility of the
  • fast_forward00:20:30 - neurons, what you get is that they decide in a way which part or which neurons
  • fast_forward00:20:36 - will be responding next,
  • fast_forward00:20:38 - which in the end will lead to the other bit action to perform.
  • fast_forward00:20:42 - So you propose a model where you say, well, you can think about this as a system
  • fast_forward00:20:49 - of competing pools of neurons that are dedicated to certain aspects of a task,
  • fast_forward00:20:55 - let's say, a goal that I pursue.
  • fast_forward00:20:59 - Then these neurons have the capability to maintain the state,
  • fast_forward00:21:03 - like a memory, like it's It's a mango leaf system.
  • fast_forward00:21:06 - And they have to calculate to rapidly switch. They have to calculate because they're by state.
  • fast_forward00:21:11 - By virtue of how they're wired, they can maintain multiple states and they can
  • fast_forward00:21:16 - switch from one to the other.
  • fast_forward00:21:18 - But now every pool is dedicated to one feature of the task, like the goal, or a cue, or an action.
  • fast_forward00:21:28 - And then you show that you could then get this switching behavior by wiring
  • fast_forward00:21:35 - these pools of neurons up in a sort of a smart way.
  • fast_forward00:21:38 - That if I get a non-specific excitation across all these pools of neurons that
  • fast_forward00:21:44 - I can now flip, that I can sort of go for QA or QB or response A or response B.
  • fast_forward00:21:49 - You're correct? Yeah. Okay, cool.
  • fast_forward00:21:53 - And you demonstrated to us that this can work.
  • fast_forward00:21:56 - But you demonstrated that in a very low-dimensional space because now we just looked at one queue.
  • fast_forward00:22:03 - But what we looked at earlier or as you discussed it earlier in your physiology,
  • fast_forward00:22:10 - you see that actually you code all the features of the desk and all their combinations
  • fast_forward00:22:13 - and it's changing over time. Thank you.
  • fast_forward00:22:17 - So how do you see that model then scale? Because every pool of neurons in your
  • fast_forward00:22:21 - model is one of these, right?
  • fast_forward00:22:24 - One of these cues or one of these cue combinations, right?
  • fast_forward00:22:29 - So do you see that scaling being feasible for your model?
  • fast_forward00:22:34 - Okay. So each pool is not exactly only one feature of the task.
  • fast_forward00:22:39 - Because the pools that I showed in the presentation are actually modules of a network.
  • fast_forward00:22:47 - And each model is composed by eight groups of excitatory neurons and one group of inhibitory neurons.
  • fast_forward00:22:54 - So each of this group is responding to one, for instance, if they are coding the goal.
  • fast_forward00:22:59 - So if it's a model dedicated to the goal, they would be coding eight different goals.
  • fast_forward00:23:04 - So each pool would be responsible for being selected for a different goal.
  • fast_forward00:23:09 - And the same happens with the other model. So inside each model,
  • fast_forward00:23:12 - we have a brain selectivity for eight different variables of the same feature.
  • fast_forward00:23:18 - So I think it's very realistic to scale it up because it will be taken with
  • fast_forward00:23:25 - the model that we presented today.
  • fast_forward00:23:26 - We already have many features covered.
  • fast_forward00:23:30 - Like if each model is actually coding the eight different variables of a goal.
  • fast_forward00:23:39 - So it's not only red and blue, like in our task, they could also code three, yellow, till eight.
  • fast_forward00:23:46 - So actually the model is already capable of explaining quite well.
  • fast_forward00:23:50 - So the scaling would become problematic when I start to add goals then.
  • fast_forward00:23:53 - So you have the grouping around goals that would be a critical feature.
  • fast_forward00:24:00 - As long as this population is linked to that goal, I can capture the whole set.
  • fast_forward00:24:07 - If there are no more than eight goals, we can explain the data. Okay.
  • fast_forward00:24:13 - That would be unbelievable, you say. Yeah.
  • fast_forward00:24:15 - But the other thing that you can think of is that your model is a labeled line kind of model, right?
  • fast_forward00:24:22 - So the synapses I get must be uniquely linked to some feature, right?
  • fast_forward00:24:30 - Otherwise, the dynamics cannot work, right?
  • fast_forward00:24:34 - So it would also mean every neuron is dedicated to just a sort of permanently
  • fast_forward00:24:39 - dedicated to some aspect of a task.
  • fast_forward00:24:43 - But isn't it an important feature of a working memory system of prefrontal cortex
  • fast_forward00:24:48 - that you can dynamically be allocated to any task?
  • fast_forward00:24:52 - So how would that work? I have my pools of neurons and now I'm doing a discrimination task, but maybe.
  • fast_forward00:25:00 - The next hour, I have to do a navigation task, or I have to do an operative
  • fast_forward00:25:04 - conditioning task, or whatever, I have to deal with other kinds of problems.
  • fast_forward00:25:07 - So how can you have this flexible allocation of the content,
  • fast_forward00:25:12 - if you want, the semantics?
  • fast_forward00:25:13 - Well, I think this model could be pretty general, because you just change your length with,
  • fast_forward00:25:19 - time, the new task, and then you change your semantic activity between the modules
  • fast_forward00:25:23 - or from the input that you get, and then you will have four different kinds
  • fast_forward00:25:28 - of behavior with the same model. Really? What?
  • fast_forward00:25:30 - So I think if we... Yeah, monkeys were really well trained in these distance discrimination tasks,
  • fast_forward00:25:36 - but if they would switch to a different task, I think this would be learned
  • fast_forward00:25:39 - in the synaptic connections of the module, and then you would have the same
  • fast_forward00:25:44 - data index that would be...
  • fast_forward00:25:48 - But actually, what we wanted to do was... I mean, the whole view that I presented
  • fast_forward00:25:53 - there was more like a truth of concept to say, okay, we have this...
  • fast_forward00:25:58 - If we have these heterogeneous neurons in the brain, as we think we have,
  • fast_forward00:26:05 - we can't really explain it as we observe.
  • fast_forward00:26:09 - So it's more like a proof of concept to say we have a brain and we have an area
  • fast_forward00:26:14 - which is full of heterogeneity in the excitability of the neuron.
  • fast_forward00:26:18 - Right. So, but now the scaling along goals, if I have goals that are competing
  • fast_forward00:26:26 - with each other, how would I account for that in this system?
  • fast_forward00:26:31 - You have opposing goals, opposing goals that are contradictory.
  • fast_forward00:26:35 - Okay, so there you would have a competition. So this is in the end,
  • fast_forward00:26:39 - each model is an attractor model.
  • fast_forward00:26:41 - So you would have within a model, you would have two pools which are.
  • fast_forward00:26:48 - Stimulated, and then they would compete. And in the end, you will have one winning
  • fast_forward00:26:52 - goal and one that leads to the final execution. Okay.
  • fast_forward00:26:57 - So in this case, indeed, so it's like an attractor model?
  • fast_forward00:27:02 - We're looking at firing rates of the model, and in some sense you're interpreting
  • fast_forward00:27:07 - or trying to predict performance and firing rate, but in the earlier work,
  • fast_forward00:27:12 - a very influential paper that you, which you were the first author,
  • fast_forward00:27:18 - you actually made the point that firing rate is not that informative about performance,
  • fast_forward00:27:23 - it's much more the variability of the responses that really matters if tasks are more complex.
  • fast_forward00:27:30 - Did you step away from that view? Are you back now more into the rate-coding view of things?
  • fast_forward00:27:38 - No, I'm very interested in studying the variability or which information we
  • fast_forward00:27:42 - can extract from the variability.
  • fast_forward00:27:44 - In this case, I didn't look… well, actually, we did something with variability,
  • fast_forward00:27:49 - but we calculated it by looking at the bars and pulses in our data.
  • fast_forward00:27:53 - Data because in the models that we presented, we have the stable models are
  • fast_forward00:27:59 - the ones which are very stable and they don't have variability or almost nothing,
  • fast_forward00:28:03 - not variability in the framework,
  • fast_forward00:28:05 - whereas the ones which are more flexible have much more variability.
  • fast_forward00:28:10 - And we tested that in the data, looking at the bursts and pauses,
  • fast_forward00:28:14 - and we found that actually we could, yeah, this prediction was right. Right.
  • fast_forward00:28:19 - What we don't have is anything with the behavior there because we would not
  • fast_forward00:28:23 - actually, I tried to look at that, but we didn't have any good.
  • fast_forward00:28:27 - In this step, the planning is that we don't have reaction time.
  • fast_forward00:28:31 - So that's very limited, because whenever the targets appear,
  • fast_forward00:28:36 - the monkey can perform the action, because he already knows what to choose.
  • fast_forward00:28:41 - So we didn't have reaction time, so we could not correlate it with variability
  • fast_forward00:28:45 - or with any other measure of difficulty.
  • fast_forward00:28:49 - So that's why this is time we didn't look at that, because I didn't have the possibilities.
  • fast_forward00:28:55 - But I think variability, and we proved that also in MOLLE, is very informative.
  • fast_forward00:29:01 - It's one of the key features of this flexible model.
  • fast_forward00:29:05 - So are you saying that are you moving away from the standard model of distribution
  • fast_forward00:29:11 - models of decision-making?
  • fast_forward00:29:12 - Because the standard view would be, look, I just integrate whatever information
  • fast_forward00:29:17 - I get, usually it's perceptionally evident, until I hit threshold,
  • fast_forward00:29:21 - or I hit threshold first, it's the winner.
  • fast_forward00:29:23 - So how are you moving away from that standard view on the decision-making process?
  • fast_forward00:29:29 - I think this standard model is sad, very nice and they are very useful to actually
  • fast_forward00:29:34 - to have some predictions about which behavior you might encounter,
  • fast_forward00:29:37 - but I think they are not so good to understand to really spending dynamics of the of a network,
  • fast_forward00:29:44 - So you there you have an intuition of what you would expect as an average activity,
  • fast_forward00:29:50 - But with the diffusion model, you cannot know what would be the individual dynamics
  • fast_forward00:29:54 - because they don't do predictions on the.
  • fast_forward00:29:56 - So, I think the models are okay, but they have their limitations,
  • fast_forward00:30:00 - so we must know how to use them.
  • fast_forward00:30:03 - So, because I'm more interested in the individual dynamics, so to understand
  • fast_forward00:30:07 - how neurons respond individually, I'm more interested in the cycling ground
  • fast_forward00:30:14 - models, because there you can really account and explain what you observe.
  • fast_forward00:30:18 - You know, it's still the proponents of the rate of the rate-coded diffusion
  • fast_forward00:30:22 - models is also the memory potential refining rate of individual cells that are
  • fast_forward00:30:29 - predictive of performance, right?
  • fast_forward00:30:31 - Because I can see that the flow that reflects integration is correlated with,
  • fast_forward00:30:38 - let's say, the direction times I get or with the accuracy that I get, right?
  • fast_forward00:30:43 - So the claim would be made.
  • fast_forward00:30:46 - Yeah, yeah, I know. But I think it's one of the, my point of view is more the
  • fast_forward00:30:51 - average of the population.
  • fast_forward00:30:52 - I don't think you can explain that with a single neuron. I know that there are
  • fast_forward00:30:56 - some people that say that you can and you see that in the individual neurons,
  • fast_forward00:30:59 - but there are also some recent papers that show that actually,
  • fast_forward00:31:03 - the ramping activity is an artifact of averaging different state activity with many neurons.
  • fast_forward00:31:10 - So I don't think you can explain the individual dynamics of that.
  • fast_forward00:31:14 - I think you can have an intuition of the main activity the data must be what it is.
  • fast_forward00:31:20 - But I could then still say, well, maybe what matters is the population response,
  • fast_forward00:31:24 - and as long as I can predict performance from the population response, I'm happy.
  • fast_forward00:31:29 - I don't care for these single cells, because they're all averaging out. Just noise.
  • fast_forward00:31:32 - Yeah, but then I would say, why do we have so many neurons? With one neuron, we wouldn't be enough.
  • fast_forward00:31:38 - I often feel like that. Yes, okay. So, yeah, I think if we have so many neurons,
  • fast_forward00:31:44 - according to so many different ways, because of something that I must be able to...
  • fast_forward00:31:50 - No, wait, don't you say I set the problem because I can say,
  • fast_forward00:31:52 - well, maybe I have all this variability in neurons so the average has some stability
  • fast_forward00:31:58 - or follows some distribution and that average is then what really drives the behavior.
  • fast_forward00:32:07 - But, yeah, you could say that but I think, for instance, in the paper that we
  • fast_forward00:32:11 - have together much more so influenced.
  • fast_forward00:32:14 - We could prove that that's not always the case. so that the inactivity is not turning everything.
  • fast_forward00:32:20 - Well, that's the other point, of course, right? That maybe the point is,
  • fast_forward00:32:23 - if you have a simple task, you have a simple code.
  • fast_forward00:32:26 - And rates are great. But if a task gets more complex, like discrimination,
  • fast_forward00:32:33 - if I ask, right, then you have to switch to a different coding model.
  • fast_forward00:32:39 - Maybe, but what is that simple? What is a simple task? Because in a way,
  • fast_forward00:32:43 - everything is complex. I mean, discrimination they should tax.
  • fast_forward00:32:48 - Well, if I have a random motion display with predominant motion in one direction,
  • fast_forward00:32:55 - I just integrate over this whole field. I just need the Monmic Integrator.
  • fast_forward00:32:59 - And okay, we'll go one direction or the other. I don't need to compare anything, I just integrate.
  • fast_forward00:33:04 - And just do reintegrating motion. So that's a simple class.
  • fast_forward00:33:07 - If I need to compare two stimuli, or I have to compare two time intervals,
  • fast_forward00:33:14 - I already demand more for my memory. you have to keep something in memory,
  • fast_forward00:33:18 - you have to compare things.
  • fast_forward00:33:20 - So this might be another argument. Yeah, but still I think we are much more optimal than that.
  • fast_forward00:33:30 - So why would you have so many neurons doing the same if you could have just one doing that?
  • fast_forward00:33:36 - Because in the end it's energy that we should not...
  • fast_forward00:33:40 - That's true redundancy or you can also argue well another perspective is of
  • fast_forward00:33:48 - course it goes into the simple task but if it's a simple task you have simple cues one cue.
  • fast_forward00:33:53 - But if you look at the world in which his brains evolve everything is ambiguous,
  • fast_forward00:33:59 - predictability is an issue the Markovian assumption doesn't hold the future
  • fast_forward00:34:04 - might be somewhat different from the past you have right so,
  • fast_forward00:34:10 - maybe if we have these highly controlled and reduced paradigms the force also
  • fast_forward00:34:15 - leads to a very reduced perspective on the complex of the system that we are
  • fast_forward00:34:19 - in so maybe just and then you are in something already moving in a different
  • fast_forward00:34:25 - direction but there's always a question do you want to still.
  • fast_forward00:34:30 - Park it inside the standard model do you want to move away from this idea that
  • fast_forward00:34:35 - firing rates integration integration, explains everything.
  • fast_forward00:34:39 - Yeah, no, I actually would like to move to, I mean, it's not that I don't agree
  • fast_forward00:34:43 - with these models, as I say, they can be useful for some things,
  • fast_forward00:34:46 - but I think they have their limitations.
  • fast_forward00:34:48 - And it's better that we open to different options and that we really try to
  • fast_forward00:34:51 - understand the digital dynamic.
  • fast_forward00:34:55 - Right. That's great. So, Ingrid, you, of course, also then delayed full disclosure. closure.
  • fast_forward00:35:03 - You've been a member of SPECS here, right? We've been working together for quite a while.
  • fast_forward00:35:09 - We mainly looked at models of the brain.
  • fast_forward00:35:13 - Then you went into neurophysiology of the monolingual primate and mycotics with all the genovesial.
  • fast_forward00:35:20 - And now you start to stand on your own legs at the Neuroscience Institute in Alicante, right?
  • fast_forward00:35:25 - So you have to make, you've made this tour now, you're sort of a young researcher,
  • fast_forward00:35:28 - you're building your career.
  • fast_forward00:35:30 - But But given your experience, what would be a Carni's law that we should follow
  • fast_forward00:35:35 - to understand the brain? Carni's law? Mm-hmm.
  • fast_forward00:35:40 - That I mean that it's fundamental
  • fast_forward00:35:44 - to use a combined experimental and theoretical approach because I think with
  • fast_forward00:35:50 - only experience we cannot understand the brain because we have only observations
  • fast_forward00:35:55 - and with observation it's okay but yeah we don't know about the insights,
  • fast_forward00:36:02 - and with the theory we can really stay that but theory alone would be also not
  • fast_forward00:36:07 - so good because because then we don't have anything to prove that what we are
  • fast_forward00:36:10 - saying is actually correct.
  • fast_forward00:36:12 - So for me and in my future life, I would like to continue combining both,
  • fast_forward00:36:17 - because I think that both are fundamental for us to advance in understanding.
  • fast_forward00:36:22 - Great. Then I will come down to Elegant a few years from now,
  • fast_forward00:36:27 - to check the state of the art in your lab and to see whether you managed to
  • fast_forward00:36:33 - really falsify or verify the key hypothesis.
  • fast_forward00:36:39 - So what's the key hypothesis that you would like to see tested in this time frame of four years?
  • fast_forward00:36:45 - One that I'm still really willing to,
  • fast_forward00:36:49 - show since I ended my PhD actually is that the variability in the firing rate
  • fast_forward00:36:57 - of the neurons are causing the uncertainty on the differences in modalities
  • fast_forward00:37:01 - and then this is read out by a second state,
  • fast_forward00:37:05 - process, which is actually computing the confidence based on this diverse uncertainty
  • fast_forward00:37:11 - that is called in the five minutes.
  • fast_forward00:37:17 - So this is the key thing where I would like to continue and to be advised.
  • fast_forward00:37:22 - And I hope that in some years from now I will have an answer to that and I can
  • fast_forward00:37:27 - really prove that it is very, very important.
  • fast_forward00:37:31 - It's not only about the filing rate and that we should start working on different things now.
  • fast_forward00:37:37 - Fantastic. Connie Marcos, thank you very much for this conversation. Thank you.
  • fast_forward00:37:45 - The CSN Podcast was produced by the Convergent Science Network of Biometrics
  • fast_forward00:37:51 - and Biohybrid Systems, a project funded by the European Sevens Research Framework Program.
  • fast_forward00:37:59 - For more interviews, recorded lectures, or upcoming conferences in the field
  • fast_forward00:38:05 - of biometrics and biohybrid systems, go to csnnetwork.eu.
  • fast_forward00:38:12 - And thank you for listening.

Be the first to leave a comment

Leave a comment

Your email address will not be published. Required fields are marked *

Convergent PozitivLinia Logo

Exploring the convergence of neuroscience, robotics, and AI through conversations with leading researchers since 2010.

A project of the Convergent Science Network Foundation.

© CSN Podcasts. Developed by IMCreativeWEBC

0%

Login to enjoy full advantages

Please login or subscribe to continue.

Go Premium!

Enjoy the full advantage of the premium access.

Stop following

Unfollow Cancel

Cancel subscription

Are you sure you want to cancel your subscription? You will lose your Premium access and stored playlists.

Go back Confirm cancellation