 Hello, welcome. It is September 14th, 2022, and we're in Active Livestream number 48.2. Welcome to the Active Inference Institute. We're a participatory online institute that is communicating, learning, and practicing applied active inference. You can find us at some of the links on the slide. This is a recorded and an archived livestream. So please provide feedback so we can improve our work. All backgrounds and perspectives are welcome here, and we'll follow a video etiquette for livestreams. Head over to activeinference.org to learn more about how to participate in learning groups and projects such as livestreams and others. Well, we are in our third discussion around this paper, Communication as Socially Extended Active Inference and Ecological Approach to Communicative Behavior from 2021 by Remy Tison and Pierre Poirier. And last week in 48.1 with the very same fellows in the conversation, we had some explorations into different aspects of the paper. Perhaps we can each say hello again, and something if we want to that we remembered from last week or that we'd like to explore today, and then we can just be right in it. So I'm Daniel, I'm in California, and I'm interested to look towards the latter half, the back nine of the paper, and see where we go, which is kind of what we do in Dot2s usually. And I just have a feeling that the exit from the slide on this one may put our communication and situatedness in a different place than where we were before. And I'll pass it to Blue. I heard the initial part then lost it, Blue. Can you hear me now? How about that? I am Blue and I'm a researcher in New Mexico, and I am excited to dig back into this paper. I think we left a lot of things to wonder about last week, and yeah, excited to get to maybe more into the meat of the paper itself as opposed to just discussing tangential ideas. And I'll pass it to Dean. Thanks, Blue. I'm Dean. I'm Karen Calgary. I'm interested in seeing whether or not the big road map that I think that this paper provides also can be explained as a big score as in a kind of a musical analogy, because I'm really interested in the interplay or the interaction between orchestration and arrangements. Yesterday there was a math stream where our guest was really good at sort of describing the arrangement part of what's going on in active inference potentially. And so I'm really kind of interested in looking at that because I think from a transmission versus ecological standpoint, we're kind of looking at the difference between mono and stereo. So I'm wondering if we should maybe look at that through that lens as well. I'll pass it to Remy. Hi, I'm Remy Tizant. I'm a researcher at UCAM in Montreal, and I'm the author of the paper. And I'm looking forward to discussing the paper more in detail and maybe go to toward the later parts of the paper and maybe talk a bit about language and linguistic communication, which is a big piece that is looming behind the paper, but it's not really that much discussed in it. But I think that's what any account of communication aims to go towards. So I hope we'll have the chance to discuss it a little bit. Let's go to language. It's what we're using, among other modalities, but we are using language and so that sounds like a great place to start. So what was in the paper regarding language and what was not in the paper regarding language? Yeah, so the idea was to open a window at the end of the paper towards language. So we talk a bit about conventional forms of communication and we know that most people believe that language is conventional. So we tried to tie the ideas that we put forward in the paper work by Ramstead in a paper of 2016 about conventional affordances. So we tried to say that linguistic communication, the conventional nature of the language comes from the conventional affordances that are generated in the communicative interaction. So that's a way to point toward language in our account. And we talked a bit about different types of acts of language. So informative or declarative and interrogative, imperative. So these kinds of distinctions we try to account for them in our framework. So these are various parts of what we consider to be language that we tried to treat in the paper, but there's a lot of things that we couldn't discuss, obviously, because it was like in the in the small section and at the end we cannot like explain all of language, even in a paper or in a book. So in a small section that would be obviously way too ambitious. What are we looking to explain or what is needed for language to be encompassed or at home or native to the ecological view rather than the transmission view or in addition to the transmission view. Remi first then others. Yeah, a point that is interesting is that the transmission view comes with a presupposition about what has to be explained by a theory of language. It comes with the idea that we have to account for the meaning of the expressions that are used first. So we have to understand what are the meanings of the words that we use and how these meanings are encoded in the brains in the heads of the of the speakers of a language. But that's misleading from the point of view that we adapt in the paper. So this it's kind of hard to say what we have to explain in order to explain language because for so long we have been in the transmission view. And so we are used to believe that in order to explain language, we have to explain these kinds of contents that language transmits. So but if we do away with these contents, it's kind of open the question of how do we account for language becomes kind of open in a new way. And I don't really know exactly what we have to account for in order to account for language because it's kind of changes our view of what language is. So I think there's a lot of open space in in in this question of how do we account for language if we do away with the transmission view. Dean or blue. Pretty god blue. So I just was looking back over the paper. I think it might be useful to maybe delineate the the specific like types of language interrogative versus imperative and like the specific the specificity is underlying each one of those right now. That would be maybe a good place to to just dive into the meat of the paper. How can we see these different types of speech speech acts as shaping the field of affordances. What is being done ecologically with these different speech behaviors. As opposed to the enabling aspect of the transmission view. I think they're fairly clear within the transmission view what these statements are about. And as you pointed out it's almost like hard to escape viewing what these speech behaviors are by the content like if not their content quite literally than what would they be about even talking about what they're about. But informative would be sharing a proposition or some reducing someone's uncertainty about the state of something in the world. I'm feeling this way or that weighs this much. And so on or other statements because indeed they're defined by the content and their structure. So ecologically I mean what might these types of speech behaviors be doing. Oh yeah these these types of acts of language are well accounted for in in the transmission view. But there are some questions that remain about the imperative form of communication because it's not clear that imperative communication can be understood as a form of transmission of content. Because we have the feeling that when we comment something or we request something we are not transmitting information or we are not aiming at changing the mental states of the person that we are talking to. We are aiming at making them do something in the world directly. So there's kind of a mismatch there it seems. So obviously the people in the transmission view will have an answer they will say that like when I have some for instance the desire or the intention that something is the case I then an imperative act of communication is my transmitting this intention or this desire to the person who then has this desire or this intention and then can bring about the state of affairs that I intend to achieve. But I'm not sure obviously you can delve in the details but I'm not sure that's exactly what we want to do when we produce an imperative act of language obviously we don't just want that the person has the intention of doing something we want the thing done in the world. So I believe that the imperative kind of communication is not really well accounted for in the transmission view and what's interesting is that in a way the imperative form of communication is like the basic form of communication in our framework because what communication is is acting through the behavior of somebody else by affecting its field of affordance. So we can go back to like the forms of animal communication that we described earlier in the paper and these are all can all be considered to be imperative forms of community or proto-imperative forms of communication. So that's like the starting point and we we we achieve these acts on the behavior of somebody else by altering their field of affordances to make them do something but how to account for informative acts of communication and an idea that we put forward in the paper is that the kind of acts of communication that we consider to be informative will be acts that aim to construct or alter a shared field of affordances in a way that is more durable than just an immediate imperative act of language. So what we are doing we are changing the layout of the the the context that we take for to be relevant to our joint action and these these changing of affordances do not have to be immediately acted upon I can say oh there's something in the in the context that's relevant it does not necessarily call for action immediately but it's something that is like in the background of our context of action and so it's changes the shared field of affordance but it's not directly soliciting an action from the interlocutor as would be the case for the imperative act of language. So in a way it's like as if in the beginning an informative act of language is like a long-term imperative or a dispositional imperative we put something in the context we say oh that's relevant we can act on it at some point but it's not necessarily immediately to be acted upon so that's how we can like put the the idea that okay we can change our context of action what we take to be relevant in the context of action and we can come back to it later when when it's relevant and act on it but it's not something that is to be acted upon immediately and so that's like the the the the difference between imperative and informative begins to to to be discerned there where an imperative is like I'm saying that this so this affordance should be acted upon immediately an informative is like I'm putting this in the context it changes the layout of affordances for instance there would be I don't know how to come up with an example but if I inform you of something I'm putting I'm changing the layout of the affordance that we take for for granted it aren't in the context of our interaction and it changes in the long term the interaction and it can like drive the interaction in various ways but it's not something to be acted directly immediately so that's the the the basic distinction and that's like the the the beginning of a distinction between informative and imperative and I can say a few words about interrogative too so as we said and communication is acting typically acting through the behavior of somebody else by by by altering its field of affordance and interrogative it's interesting because it's exactly the distinction between between imperative and interrogative is really similar to the distinction between pragmatic value and epistemic value in active inference so for instance when we produce an action that aims to maximize a pragmatic value we are trying to for instance to to make the state of the world correspond to a prediction of our derivative model and when we make something which is guided by epistemic value what we try to do is we're trying to reduce uncertainty uncertainty about our our niche for instance but so in a way so for instance when I'm eating something I'm I'm maximizing my my my pragmatic value because I'm for instance reducing a prediction error that I'm I'm angry and I want to so that's like a basic example so but when I'm exploring my environment when I'm making a visual decade to explore my niche so that would be an epistemic action so to speak so interrogative acts of language are exactly analog to epistemic actions but there are epistemic actions through the behavior of somebody else so instead of for instance by myself exploring the niche exploring the field of affordance I can produce this act that will make the other person this close to me an element of the field of affordance that can then reduce my uncertainty about what is the context of our action so that's really the distinction between imperative and interrogative is really to be understood along the line of the distinction between pragmatic value and epistemic value so that's the beginning to an introduction to these these distinctions in the in the types of acts of language thanks to me you totally you totally read my mind there about like what I wanted to to elaborate on even though my question was like very poorly framed um yeah I think that that's very useful for especially in like the goal directed aspect of imperative versus interrogative or informative like what how they interact with our maybe or we don't have a joint goal versus we do have a joint goal or the goal is to develop some shared state of a shared mental state so that was really helpful thank you another note is that the informative behavior is like the the base class of the transmission view because it seems to be like it's updating one on the details like what if I'm saying there is a gallon of milk in the fridge truly what else could be conveyed other than the content so it seems like the most natural and then there are specific kinds of content like um sharing the content that you have a question or sharing the content that you would like somebody to act and so in that way as you described many forms of speech behavior are well accounted for in the transmission view and that accounting is actually kind of a callback to that math stream for with David Spivak about accounting systems so these are like accounting models um and then you highlighted that the imperative in the inactive and ecological view on communication that the imperative is perhaps the basic form of communication and that can be seen developmentally through a child and the kinds of communications that they might engage in as well as from an evolutionary perspective with the kinds of audio or chemical or physical stigmergy and marks that different kinds of creatures used to communicate within and across species so those are action oriented communications as we can then say all action oriented communication or all communication we can say is action oriented in and out this way so then imperative becomes the base class shaping the field of affordances if that's what the ecological view on communication is is going to be most natural with kinds of expressions that we can say are imperative and then informative updates are like one step removed from direct action requests so saying turn the handle to the left is an imperative if you said I'm going to inform you about two things the handle can be turned to the left and if this happens you may want to turn the handle those are breaking down a capacity for action into sequential or modular or recomposable statements but those statements are like hiding the consequences of action and so in that way we we can have an action oriented view on communication and then see different kinds of other behavior as like modified or truncated or censored or hidden action modification and that may or may not entail needing to even explain how meaning is represented transmitted or received yes Remy oh yeah and that's exactly it so the that's really the the idea you explained it very well and another example that we that we often use which here is the the joint action of cooking a meal with various people so like an example we used in the last discussion the example of moving a couch so another example maybe to change a bit our intuitions about communication so when people cooking a meal together we can suppose that some people some of the so suppose that we're in the kitchen of one of the the person and the other doesn't know exactly where all the things are and so the person could say there's flour in the in the in the cupboard or in the pantry and saying that that's an informative active language but that could that's like that might be a hidden imperative in so far as at some point in the interaction in the cooking of the meal the flour might be needed and at this point this alteration of the field of affordances this appearance of the flour as an affordance and in the context might drive the action of the person when the flour is needed at a certain point the person will have this affordance in its field and will go to to to seek it if it's necessary at that point so the informative is as I said something like a dispositional or long-term imperative something of a hidden imperative that can drive behavior in the long term so so it's like at the beginning this distinction is more this a distinction of like is this affordance to be acted upon immediately by you or is it posed in the context and can like alter the dynamic without being directly acted upon immediately so that's how the basic distinction is drawn and yeah I I think that that's that's how we try to to account for informative without describing it as transmitting a content or an information so that's that's the idea and I think that you explained it very well in your in your in your in your yeah in your explanation good yeah it's very interesting it's just another consequence of centering action so many things that the pragmatic turn pragmatism earlier than the so-called pragmatic turn in the neurosciences and all of the developments in cognitive science in many cases it's been like an information or an inference view yielding to an action view and that has just played out differently on different frontiers and it feels like in a way language is like the stronghold of information views because it's like so amenable to syntactic and discourse analysis in a and compositionality and all these features that we see in formal models and in computer systems that it feels like oh maybe all the funny stuff that people do and say we'll figure that one out too blue and then Dean so I just wonder about like mimicry which is something that we see like in the animal kingdom a lot and like whether that is imperative I mean sometimes I guess it can be imperative like um watch me do this and then you do it but but not always so sometimes it's like um you know an animal will see another animal do something and then they try it too right and and so what um I don't know I just don't know how to like in the terms of teaching and learning and like shared goal directedness like what is maybe the role of mimicry and how does that like what does the effect on the field of affordances does it just open up like you see someone do something and then your field of affordances opens to that is also possible for me or does that not have maybe like a semantic analog I don't know yeah that's a really good question and often people who who adopt the transmission view will adopt the view that the the main driver of the evolution of language in human societies was the transmitting of knowledge and of skills for instance and of various types of activities so that's like a a clear example where it seems that the transmission view has a story to say when if we if we take for granted that like imitating somebody or learning something to somebody is in effect transmitting a knowledge or a certain kind of of mental state or cognitive state so that there's clearly something interesting there for the transmission view but I think that an alternative possible analysis you you just heard added is the idea that you what when you're interacting with somebody and specifically with somebody of your of your own species seeing this person do something can open new affordances for yourself for instance if I say I see somebody that looks like me do something I'm I have learned that when there's somebody like look looks like me does something usually maybe with a little bit of effort I can do it too so do somebody doing something is a way of opening opening up new affordances for me and trying to achieve it so in in many cases mimicry will not necessarily be communicative that is because an organism can do something without having no is not communicating nothing to anybody but it happens that somebody else watches it and imitates it so in that case there's not necessarily a communication but in cases where there is communication involved I think we can analyze it analyze it just as you said that is saying that the action of the organism opens up a new affordance for the the the person who's watching it doing it and usually in cases of explicit skill learning for instance the the the expert will like do the thing more deliberately more slowly more explicitly to like show what are the relevant movement that has to be made in order to achieve so that's like making more silly in the kind of of of movement that has to be made to to succeed in in performing the skill so I think that's a way that we can understand this kind of skill learning in the transformation of the affordances that are available to the person that's receiving the the communication and that's how I would I think I would analyze it one simple example kind of building on this making certain actions deliberate then Dean interested for your thoughts even showing something deliberately verbally calling attention or just the position of the body for a physical skill it shapes the field of affordances specifically where somebody can look so if you're like okay I'm gonna swing the bat and then now I'm gonna show you what my hand you know this isn't usually how I do it but I'm gonna show you how I was holding it at that time you have shaped their field of ocular motor affordances in and that is what enables their generative model to update and so nowhere in that account is I'm transmitting that this is to be held between these two fingers let alone counterfactuals like here's what you do if your hand is slippery but by bodily posture including vocal box bodily posture we can have an integrated account of mimicry skill transfer again transfers playing way too strongly into a transmission view but skill resonance or something similar that helps recognize the cognitive entities learning process Dean um okay well I want to take what you're saying but I'm gonna put a little bit of a weird flex on it because I want to try to combine blue's pin and Remy's pin and Daniel's pin into one kind of question that comes back to the paper so like the last five minutes of the conversation has been focused on the act of and as typing when it pertains to language and communication and and blue bringing mimicry in is a is a form of as versus a lot of the emphasis in the paper which is on the of of communication and language so here's my question I went back and read the paper again and I caught myself asking what do I hear when I'm reading something or I'm listening to someone what do I hear as both processor and the producer as we mentioned off camera and a ranger of something being signaled which I'm picking up on and decoding so rather than just leave it there I'll park that for a second because I think that's the the the nut of what I want to ask as a processor and a producer as an arranger I can hear things in mono I can hear things in stereo I can even hear things if I go to my apple music in spatial audio so that is a field of affordances question as opposed to a source of affordances question and if I only limit myself to the source I'm going to be asking only of type questions right so if I want to stay in the as of if I want to be able to hold up two things at once at a minimum the stereo or the spatial that's when I get into the ecological even if it's specific ecological because I'm not collapsing down to a single source as my point of contact which then asks leads me to ask the fact that communication can be formed as an of types meaning mono stereo spatial mimicry interrogation the fact that it can be broken out into those things does that introduce self-production into the self-processing in your paper because I know there's a heavy emphasis on the of on the processing but I wonder if you keep the as when blue brings it up or when daniel brings it up if we're if we're keeping it on that even a specific ecological level are we not when we hear something producers as well what are we producing I have no idea but I know I hear something even when I read what you're typing oh I'm not just hearing the tappity tap I'm hearing something we are prod produce we are produced like it's changing I'm hearing that not reading that so if I incorporate that what am I doing and I'm not the expert on communication so I wonder what the expert can tell me about them yeah that's a good question um yeah I'm not sure what to say about that we'll first and then we'll continue so I wonder if what you're hearing Dean has anything to do with the internal monologue right so like as you are reading something then you are also hearing something right and so it's like there's this internal I mean like I have a very I'm a talker person and so I like I have a strong internal monologue like all the time running like even getting ready in the morning okay I need to get my shoes and get my purse and my keys and I have to pack my lunch and like all these things right like that happen so I'm constantly hearing that like as if I'm saying it maybe to myself even though I'm not speaking but but I wonder what um you know so I definitely am a producer of whatever I am reading like I reproduce it I reproduce it though right so I reproduce it in my own mind my own internal monologue but I wonder like I know not everyone has this internal monologue and so so what is it then the same for the people that don't perhaps have this internal monologue I mean maybe they carry around a picture of all the things that they need to get ready in the morning so yeah Justine yeah and I think you bring up a good point blue is it a monologue is it a stereologue is it a quadrophonic log for some people I mean I mean there was an availability way back when maybe even before some of you were born there was quadrophonic sound I remember the who album and quadrophonic sound and like to my simple ears I didn't understand what that even meant just because they were able to break it out into four parts didn't didn't change what I was both producing and processing so but but the fact that we can type it like that like not to type it out but literally turn it into types is the part where I don't really have the background to be able to explain how that that ability that field of affordances as opposed to examining this as a single source of affordance changes the the communication game because it certainly does speak to mimicry and what we can and cannot pick up it certainly speaks to Daniel's point about about what happens when when layers appear and disappear because we narrow down what we're picking up on versus have both of our ears working even though we think it's our eyes that are doing all this skating right so again I'm not I'm not trying to ask the impossible question I'm just curious in the sense of producer and processor and what that arrangement is dependent upon what the what the because Daniel brought it up in the last lecture circumstances matter so but as a hearer I wonder I wonder if we if we know enough about ourselves to really be able to to to give this anything more than as as Rami kindly pointed out in the paper give this more than sort of a big roadmap view if we if we are taking context into consideration so thank you Rami yeah I think the discussion about internal monologue is really interesting and there's a lot of I think that you said the Dean that are I think really interesting but I want to go back to the the idea of the the monologue and there are some people who believe that like Chomsky who believe that language is not even for communication it's about structuring thought and so there's there are people who believe that the language is first and foremost something that is a property of the structure of our thoughts and not a property of systems of communication and so there's this idea that's like the the the the inner monologue comes first and then comes the communication of this inner monologue in the in in communication so we have this basic that is once once more in the rent in the transmission view that we have our train of thought we are we have our that the the inner speech that we that we have in our heads and then we translate this inner speech in over speech that is then taken by somebody else and I think that this gets the the the the the priority wrong between outer speech and inner speech and there's been a lot of work in philosophy by notably Wilfrid Sellars and by others in psychology like Vygotsky who say that the first form of speech is over speech that is then internalized in in that becomes like that becomes inner but first in the beginning there's only over speech a child that learns language does not have an inner speech there's only outer speech that is expressed directly that the child speaks the child do things with his words in its environment and does not directly automatically as the corresponding inner monologue in its head and there's a process that's really interesting I think that is where we see that beginning to happen where the child speaks to itself alone when there's nobody else around and so there's this transition where language is used in communicative interaction and at some point the child use a language alone and there's a question of why why does it do and I think that our view of communication is really interestingly linked with views of of inner monologue as self-queuing or self-regulation and so the child can auto regulate its field of affordance when it's alone when there's nobody else around and in time this inner the outer for instance this outer monologue but alone monologue becomes overt because the child learns to not necessarily produce the sound that correspond to the producing of these acts and there's there's this integration that there's this internalization of outer discourse that becomes inner discourse and there's a lot of cute anecdotes about that my nephew is I have a nephew the son of my my sister and he's like when he's about to go to sleep he's in his bed and every night he's like rehearsing the events that happen in the day and he's saying like the the the comments that that the rules the norms that he heard in the day oh I don't have to do that I must do that and so he's like rehearsing what happened in the day so that this outer monologue when he's alone in his bed is like the beginning of inner thought where we're seeing his thoughts out in the open and at some point this inner this outer monologue will become internalized and we won't hear the his thoughts out in the open but I think that's like the process that it takes and there's also the idea another point that we for instance when we read usually we read silently so we read what's written and we have the the corresponding inner monologue but it's known that in ancient times when people were reading usually they were reading automatically allowed loudly and they it's it took a training in order to be able to read without speaking the words and so so like in in middle age and in I'm not perfectly sure it's it's it's it's it's terribly accurately we we should verify it but I I believe that's that's true and it seems that I heard that so it seems that first there's the outer monologue the outer dialogue and then there's the outer monologue and then there's the inner monologue that comes after that and I think that's more the the the true direction of priority between inner monologue and outer monologue so I just have a cute anecdote about learning learning to talk so I have two kids and they're super different my son did as your nephew did Remy like it would be in his crib talking to himself talking to his stuffed animals and kind of recreating events right right in a in a similar way he also babbled a lot like a lot of like this is just always like just saying nothing from the time he was Chinese and has an incredible vocabulary I mean now he's six but people are like wow how old is this kid I mean he's reading at three and just that is like very very verbal and my my daughter is super different so she did not really babble but she signed like crazy and they both went to the same school they both teach them the same signs like to have say milk and cracker and cheese and you know bathroom and all the things right so so like they both went to the same day care same niche environment and she picked up the signing much more than my son ever did he used maybe one or two signs she's like 20 like I had to go learn the signs and before she could sign she would just like stand in the kitchen and scream like just unable to communicate whatever it was that she wanted and it's so odd like they're they're just like two totally different kids but she when she was learning Chinese I had a parent teacher conference with the tea and like they've both been learning Chinese now they're in private school they've been learning for like eight years so when she was first starting to learn she wouldn't speak in class and the teacher and I had a conference and the teacher said like I am concerned about your daughter she doesn't participate in class like she doesn't have to take Chinese it's not mandatory if you wanted to put her in a different class so she could have a silent reading period and I was like oh so it occurred to me like and I knew that she and because she didn't babble and like I was somebody handed her a lollipop she was like too and she I said say thank you and she did the sign for it but but I mean she really did not speak like would not form clear words unless she knew she could say them perfectly and so I centered a tutoring with the Chinese tutor like just twice three times and then it began like so she was like listening can I repeat this can I engage with this word correctly and and she just was very more much more careful in that but both kids as soon as they could talk talked in their sleep like crazy anyway there's my cute story nice stories the development of the self-queuing and self-speaking is very interesting even earlier on in behavioral analysis and behaviorism like in BF Skinner's work there's extensive exploration of how covert monologues and covert speech and societies of mind and then I guess more recently it's become better understood how electrical activity in the body and in the brain are still initiated during covert speech like it's not an audible speech but actually in various ways it still is embodied and then this question of reading perhaps is another area where we're talking about in this language area what is similar or different with listening to speech participating in speech audibly is it different if it's red or braille how does the modality and the media influence some of these things Dean um again I I think the proportions of two ears to one mouth is something that we need to kind of really hold on to in this idea of single source and monologue versus what potentially is a stereologue in a field of affordances as opposed to a single source I don't know why why we've been I don't know what advantages are given by having as an ability to hear in stereo but they must be better than having a single ear or a single receipt or a single reception site and yet we we don't have two mouths so I again I think it's understandable why from a transmission point of view the deliverer standpoint we always collapse back to a source a single source but if we flip that and now think about it from the receiving standpoint how does that how does that change our sense of what communication is because it doesn't change if I cover one of my ears or if I have an earbud in and it cancels and yet something fundamentally must change in terms of communication and the joint action even if it's monologue in my own head versus dialogue or trial log or quadrilog now again I haven't done any research to try to try to figure out what other people have done in terms of looking at that but to your point though Renee in terms of let's get let's even get past transmission we're getting trying to get past the idea of what is being delivered and being and trying to gain a deeper appreciation of what is being heard so blue I'll tell you a little story I have a loquacious son he's in his 30s now but he's actually as he's gotten smarter and wiser speaks less like occasionally he gets out of developing word silence that could serve an entire community and I find that really fascinating I don't have to comment on it it's just you notice that as people know more they maybe they sign more maybe they find other ways to to make their point and it goes back to that timing your take then so yeah great um there's so many references like the true way is the one that can't be spoken of or those who do do and those who can't teach all these kinds of funny um you know poking fun at different archetypes and so on and how that's related to wisdom the two ears and one mouth which you've been referencing throughout the conversations and the the quadrophonic sound quadrophenia I'm sure was amazing um it is true that the enrichment and the spatial nature of heard sounds soundscapes is uh it's like something that can't be turned off like the receiver all of our receiving is bilateral visual nostrils and smell proprioceptive is is heavily multilateral it's but it's it has a coarse bilateral symmetry but um the mouth as our action states needs generative uh cohesion perhaps just a just so story but I don't know of any cases where there are um now there are some bodily sounds that are bilateral like perhaps the um vibration of a cricket so it's not that like sound production can't be bilateral it's just that wherever sound production is localized on the sender when we when we highlight content and its transmission where does that account begin what has to begin with the content and the representation in the sender and then you could ask whether it was um fully or partially or deceptively recognized by the recipients potentially when we censor the recipients action but even calling them a recipient kind of puts them onto a back foot we're just centering the agents engagement ecologically and they're interfacing with fields of affordances then we we do see that stereophonic enrichment of everything and no matter how narrow or um how many Shannon bits or Kolmogorov bits it doesn't really matter because it's going to be unpacked in a generative model that brings in aspects that aren't and don't need to be in the signal and hence aren't or don't need to be in our account of the signaling just think i'm trying to bring back the inference into the active inference here because the active part is definitely the transmission piece so you know thank you for for reinforcing that's but i i do i do wonder when we do bring the inference back why are our sensing affordances are those bilateral forms there must be some advantage to that and so is it as blue pointed out because of ads not just the heavy emphasis on the of what what was the product of my transmission how am i pulling in this communication and what advantages are there to not getting strung out or hung up on the mono log the single source one thing that makes me think about is in caricature disinfo representations of bug vision insect vision it will look like they're seeing many separate cells whereas if one were just to take a breath and think about their own visual experience we don't see two fields of you so it's like not really making sense why an insect would see multiple so the visual field because there's quote objects out there we have all these different ways that we can talk about how we know that we're experiencing an integrated generative model like the blind spot the differences in visual acuity in the peripheral vision color vision in the periphery ocular motor saccades and the way the attention is suppressed during saccading so that it's not causing us nausea all these features are really important and they point beyond a photon transmission view a vision into this generative stereophonic vision and then in the audio setting having multiple oral orifices and also even covering them up like there's still transduction through the skull and so on we are also seeking to flesh out a generative model of sounds in our environment so if we cover our ears like that proprioception will help us understand that even if something sounds kind of quiet we actually we can recognize that it's still a loud noise or if we have one ear in a pillow we could recognize that something is happening even though if it is being heard differently on both sides um so having multiple sensors is very compatible with a unified generative model experience because it forces one to abstract beyond any sensor reading if you have one midline thermometer then that's your point estimate for temperature or you could have like a Gaussian blur like I'm just less confident about the temperature further from my one central sensor but maybe I'm maybe it's a tight Gaussian maybe it's very broad but this is the best estimate why would I have any other estimate centered anywhere else when you have two thermometers now there's a generative modeling question about whether it's going to be a Gaussian with a peak in the middle or whether you're going to have a peak more on one side and have it just go through the other depending on those sensor readings but at that point you're ending up with some distributions which if we think of that distribution as like what is experienced as like maybe the product of the sensing process that distributions enrichment is of a different type and yet of a different simplicity than any number of sensor readings in terms of what their raw information are so it seems very relevant this two ears one mouth not just for listening twice as much as we speak but for listening twice as actively as we speak or something like that what do you think about sensors Remy? Yeah I think that you explained it well it's yeah I think it's useful as you said to have multiple sensory channels that can be active because I think that's an argument from how we in its book in 2013 book on the predictive the predictive mind I think he says that if you have a tribunal and you're trying to establish the for instance the culpability of a convict you hope that you have multiple independent sources of information because if these sources of information are influenced each other you have a less reliable I would say I forget the word in English so I forget the word but you have a less reliable information about the the case so like it's useful from a from a reliability standpoint to have multiple independent source of information so that that may be an explanation of why we have that may be an explanation of their also their relative modularity of perception so there are we know that there are great limits to that modularity but there is a certain degree of independence in the different sensory modalities that we have so yeah about the two here is obviously I think maybe if we had if our auditory sensory system was only devised to communicative interaction maybe it would be okay to just have one here one here because we could like we know who's talking to us and so we we can direct our our sensory apparatus toward this communicative signal but because we used our ears not just to to attune to a communicative interaction but also to know what's in our surrounding it's really clearly more useful to have like to to have a great reception from from information from anywhere around us so I maybe if if the audition was only about communication it would be okay to have just one here but because we want to know as precisely what's happening there and at the same time know what's happening there maybe it's it's it's it has been more useful to have these two ears two ears on both sides so so maybe that there's something about that yeah one short note and then Dean is there are also animals with an anatomy so that they can direct their ears which is seemingly very analogous to the ocular motor succading it's like an audio succade and it just the human form outside of the ear wiggling which is maybe slightly vestigial representation of that but we have opted for a strategy with fixed ear positioning but a mobile neck which also makes succades so that's kind of our low speed audio um succating Dean yeah I think the only advantage I would see it having two mouths is that I could eat the french fries and drink the milkshake at the same time and I don't know what that would be like it was just overpowering but what I think is interesting in the dot dot one um blue brought up the idea of so how do we make sure that there's a difference between content and context and I think whether it's two eyes two ears two nostrils whatever there must be some advantage to the form because as I said even if you cover one of those intakes um it's not just a redundancy issue it's a recursion issue it creates that ability to see context hear context and smell context I don't believe we would have those things and as I said I think I'd get overwhelmed with context of having both french fries and milkshake it would just blow me up so but that that's just that's just this little self thing do not give people two mouths it could be overpowering one aspect might return us to the question of compositionality it seems like visual composition and temporal composition in music and in um heard speech as um speaking in terms of generative models have highly factorizable or sparsely connected generative models famously connected to like the where and the what streams in visual neuroanatomy and that's been explored in some active inference papers with awareness and whatness as being factorizable in the generative model because it really is the case that different wares can host different what's and so it makes sense to like have those two totally separate um taste and smell which are um they're more mid-lined though there are two nostrils I'm not totally sure on on whether there's like you know one or two smell chambers or anything composition of chemical bouquets seems quite different like two paintings that you like next to each other or two songs in succession that you like or it's a little bit different than two smells or tastes sequentially or composed um and so maybe these more sometimes called basal sensory or cognitive features are less effable and this kind of relates to a a cognitive stack where verbal and symbolic communication is like the tip of the iceberg supported by like biological processes that are differently or less effable and at first only pointed towards like through metaphors and speech butterflies in my stomach and so on but then ultimately entirely inexpressible maybe with some stopping between like being able to express through the body like what it's like to do a certain bodily position um or have a certain experience or be in a certain physiological state um and this kind of casts a light on the transmission view complimenting the ecological view the transmission view is like down propagating symbolic communication and seeking for that to apply all the way on down whereas a more bottom up and arguably eco-evo-devo informed perspective is grounded in the ecological setting and then allows for the crystallization of symbolic processes including communicated symbolic processes which aren't aren't they supposed to be the stronghold of the transmission view but then if even that phenomena can be captured within an ecological frame then there's a extremely strong basis for the the integrative cohesion of an ecological view as opposed to an extremely piecemeal and um question begging transmission view like we talked about semiosis and biosemiosis in live stream 47 so will we see um verbal communication and behavior as something that is enabled and with analogy but not the same as cellular transduction or are we going to seek a cellular transduction explanation based on verbal behavior let's return to some of our topics all right few areas to go blue any thoughts just anywhere you want yeah so i just wanted to follow up on what you um said and i just think that the beauty of the ecological approach to um understanding communication is that communication between cells be it semiotic representation or other and between animals maybe with gestures maybe interspecies communication like you know you tell your cat come here and the cat comes or dog or whatever so i think that that the ability to model systems like all systems it is very um important like if if the transmission view is not capable of representing all of the situations from cell to cell communication to human to dog communication and and all along the the line um i i think that maybe uh it's in the way that the FEP and active inference is scale free and and like doesn't have like i mean it's it's across any hierarchical temporal cognitive scale i think that that the way that we view communication also needs to be broadly applicable yes ramy so yeah that's that that's the aim but clearly some people will say that uh sophisticated linguistic communication cannot be accounted for in terms of in terms of ecological communication because as daniel said the symbolic communication characteristic of language seems to be too removed from the the kind of basic interaction described by the the ecological approach so there's this like there's a lot of work to be done still to to really explain these the the high level linguistic interactions that we have in or in in terms of of ecological inactive communication but yeah there's another example of of communication as you said we can account to maybe for a communicative interaction between cells but there's also there's been work on communication between between plants and how plants perceive their their environment in terms of affordance there's been work on that and so i think that there's a it's just to say that there's an it's another another way in which we can develop the the ecological approach and it seems that that if we yeah any any any organism that organism that behaves so any organism will have a field of affordance and it seems that if it wants to act on other organisms in its environment it will act in affecting its field of affordance and it looks like the kind of communication that is that play between trees for instance which alerts each other about for instance the many threats in in the in the forest or sources of of useful resources in the in the ground so these kinds of modifications of of the of the yeah of the i don't know how exactly the mechanisms by which these these these these these communication take place but we can also see them as modifications of the of the field of affordance of trees so i think that that was when i saw that i was really i thought it was really cool that the trees behaved at a really really slower time scale obviously but like behave when they go when they they send like their their roots in certain directions and things like that we can see it as them exploring their environment and reacting to a field of affordance that that they so obviously that that's also controversial because there there's all this debate about the cognitive states of of plants and do they have a consciousness or all that thing but i think that we can see that they probably have something like a field of affordance and it seems that they can communicate with each other via these modifications of their field of affordance without us having to see that they are transmitting content which would be really weird from a standpoint of a plant thank you blue so just i like if you haven't seen the plant robot created by the mit media lab like check it out because the plant like they connected it to a set of wheels and a motor and literally will wheel itself back and forth from window to window like searching for sunlight so i mean when you modify the field of affordances of a plant it will it will take advantage of those affordances for movement or or any other kind of thing and as to the the complexity of like linguistic communication okay but like also like the complexity of cellular communication especially in like developmental biology and all of the growth factors and signaling cues and i mean it's highly complex and like the recombinatorial aspect is also incredibly crazy like i i think it would take a lot to convince me that linguistic communication is more sophisticated or more advanced than biological communication in a developing organism i've not sold on that thanks team ruby can you can you help me understand a little bit more about the um the influence piece uh in the joint action if if i'm in madrid um a lot of people who are trying to get my attention i'm not very i'm not i don't speak Spanish very well just barely get by and so a lot of times things are spoken to me and there is a there seems to be a break in terms of the amount of sensing i get and so that joint action could could you maybe especially in the light of that earlier uh congress part where you talked about young kids they think they it appears they're talking to themselves but they're talking to the world and maybe they're not influencing the world but they would like to could you maybe talk about that sort of sensing side of that joint action yeah so i i think that's what uh kids that talk alone are doing they're not they they learn to talk by talking to each other and affecting other persons but when they learn to talk to each other what they're doing they're they're influencing themselves i think that that's the part about self-cuing so what they're doing they're they're modifying their own field of affordances with their with their this this inner speech and that's how they they they learn to self-cue and self-regulate themselves by bringing some some images in mind or other things like that so that's the i think that's what happens when when a child learns to to speak alone and there's obviously a lot more to say about that um when so the the question of of uh of different languages when you're in another country and you don't uh so you you you're spoken to and you you're not understanding correctly but you you you pick up some some parts of what's said so i think that um i think that the the learning of a language is often is often understood as the as the as the learning a code so learning a set of associations between between meanings and between forms that that can be either written or spoken but i think that in the the approach that we're we are advancing here learning a language is really more like learning learning a skill so you're learning to to speak and learning to understand and interact correctly in a given language as you're learning to do anything else you're learning to play a game you're learning to like play hockey so or play soccer or whatever it's the is the the same kind of learning and you're learning to interact in that kind of game and probably there's a similarity between various languages um uh because it's there's there's a similarity there's the yeah like like very i would i could talk about Wittgenstein once once again but like there's this the various languages uh resemble each other because they have this this family resemblance so because you have this familiarity with the speaking in a language even when you're uh in another place where you don't understand the language you have a basic grasp of some of the structures of the game that is being played and so from the context you can get uh understand some of the things that are said understand some of the things that are expected of you uh because you have this grasp of what a kind of game is played when we speak a language and so even if you're in in in another country when you don't speak the language from the context that in which you find yourself you may be able to piece some things together obviously if you're like just in a place with absolutely no context you're in a white room completely empty and somebody's talking to you you have no context probably you won't understand anything but if you're in a context where you can understand that some kinds of there's some some goals that are that are in place and some people around you are trying to achieve some goal you can apply your understanding of the game that you play in your own language to this new language and so you can get some things you can understand some things about what's what's what you have to do and uh how you should behave and how you can respond but I think that that's a better way of understanding the kind of limited understanding that you have in these kind of of interactions rather than in you having a code of your language and then you try to apply this code in this you I think that that the the idea of learning a skill learning how to play a game is a better analogy to understanding how you you come to learn a language and how you come to apply this this knowledge in other contexts where the language that is spoken is not the same that you're mastering does that kind of confirm that it couldn't be content first that it would have to be context bound yeah in my view the the context is a central element in in in any communicative interaction in any cooperative communicative interaction I think it's the in the context what what I I call the context in the paper appears as a shared field of affordance so that that's the the affordance that are present in the in the in the the environmental context in which you find yourself in with the other person and that that's that's this element that often pre-exists the communicative interaction is there a great great source of constraint on the the communicative interaction so yeah when you when you you come to speak with somebody you're almost never like in a absolutely context free place you're always in a restaurant in in I don't know in the bus in in at the grocery store you're somewhere and there's this basic set of affordances that are shared that you understand that you can do this you cannot do that and this this set of constraint offered by this shared field of affordances really drives and constrains how the communicative interaction will unfold so that the I think that even if you're in another country often the the context will be hard to pick up but some sometimes there will be shared context that you can get a grasp on and that you can that can favor that can lead to a sorry I'm losing my words that I can lead to an interaction that is that is optimal I brought it in the physical Rosetta stone as a special case of low context asynchronous communication one can imagine there would be a million different contexts by which the analysis of the stone could have been accelerated such as a person reading any of them the rules or descriptions being enacted or a labeled image or any number of contexts that we would use to interpret and act upon the texts that we read in this setting where we were historically removed from the creation then there still was residual context and it turned out that that was enough to interpret the symbolic materials but this is not like the base communication case that we then need to see all communication and all learning and education of languages and active inference ontology as being like a Rosetta stone like the definitions of words are not codes to crack they're like inside jokes or ways of communicating on a sports team that have to do with the context of the setting and asynchronous modifications in the environment especially symbolic asynchronous modifications which are relatively new potentially existing in their modern form for several hundreds to several thousands of years depending on how we think about it symbolic asynchronous modification is a low context communication whereas even the interpretance of asynchronous symbolic communication still uses and embodies their context in the perception so we know that we have stereophonic communication just because we're stereophonic even if we are getting just a Morse code symbol but it is a non-controversial here but interestingly it sounds like elsewhere that we would want to think about communication broadly as being action oriented and embedded rather than kind of like reverse engineered from the kinds of artifacts that we have today with quite literally computer codes and other kinds of you know Turing computers and so on okay I think that you often explain the ideas that the ideas that I'm trying to convey really better than me Daniel I'm not really fluent in English so thank you for explaining better than me the the ideas that I'm trying to say thank you um I appreciate it and you wrote the paper and so it's unsurprising it seems novel and I couldn't say it in any other way even when you speak English Daniel can still like speak better than than you communicate more effectively the ideas that you're trying to convey even when English is your first language well context is non-pungible so I don't think that there can be a best or a final communication it's just like a best or a final species or state for an ecosystem to be in and I guess in the last 25 minutes or however now that our ecosystem has been modified however it has been in these conversations and through reading and opening up threads uh how will it shape our field of affordances and how will we stream and play differently Dean so I'm curious from all from all three of you as a strategy um if you want to get past monologue what do you do how many times do you play playing play so as an example of like a real pertinent one right now the way that the live streams about have been set up have always been sort of a pre-brief an introduction with the authors or a play and then a a play a debrief piece to this and so the communication is never left as a as a one-off exercise um I just read the paper there's there's there is affordances set up to interact with information to be able to receive it at multiple different time points because I think in the dot one I mean you said we could look at it as slices of time but that's probably not accurate in terms of of the dynamic that's in play so one of the things the actor conference institute also does is has a textbook group now I've not been a part of the text group book group and I don't know whether they do it in sets of three as well to create that kind of triangulation so that the interaction with the communication doesn't become monolithic or monistic or here here's here are the words that have come down now off the mount accept them and start start carrying them out even though you don't have a context right chiseling them into the stone doesn't increase their priority at least if you don't have the context so what I'm curious about from the three of you is if you want to get past instructionalism in terms or if you want to get past transmission is there a basic requirement to have multiple reinterpretations or at least the opportunity to reinterpret like does in from a communication standpoint do we must we have reflection and time for that do we have to have time built in for that or are we are we then bound to a transmitted state on anybody can answer yes blue first so not like in the way that we have reflection like we have a pre brief and then the actual brief and then and then the reflection time in the live stream series so so not in that way but but I do think that in in effective communication there is like reflection as an essential component of that like communication when you have a conversation it shouldn't be about just waiting for your turn to talk right I mean there should be some reflection of what the other person is saying in what you're saying back to them like hello how are you doing today I want scrambled eggs like if that was a conversation it wouldn't like be it wouldn't work right like so there has to be some element of reflecting back what the other person is saying to you in in communication that's effective it's not just this like tower of disjointed blocks being built yes Remy yeah I think I think that what you said the blue is interesting the really important element of cooperative communication is that when you're talking to somebody when you're in a dialogue you have to concede a part of your autonomy you have to accept to be regulated by the other person and that's I think where some people who like talk past each other and don't listen to each other and directly that happens really often some friends talk talk like it seems like when when they're waiting they're just waiting when they're listening they're just waiting to say what they want to say so these people don't like don't respect the kind of pragmatic principle of like you have to pay attention to what the other person is saying and to have a proper to have a correct coming at even traction you have to concede the like you have to accept that your field of affordance is regulated by the person who's talking to you you have to let yourself be influenced be constrained by the by the person who's talking to you just because you you have to listen to it and you have to refrain from like turning away and go go go do something else you have to like it's it's really just accepting conceding a part of your autonomy and that's another idea from a deep hollow as an active approach of language where like there's in communicative interact in any cooperative interaction but in communicative interaction particular you have to like in a bit a part of your field of affordance to participate in this interaction and listen because suppose that what I'm saying is really boring to you and you just want to go eat your piece of cake in the in the fridge like you have to in order to participate to maintain the interaction you have to inhibit this this afforded solicitation and to let me regulate your your your field of affordance and so there's this part of like I concede I concede something in order to maintain the interaction and when there's no no more interest or I'm not able to I don't want to be regulated and there's another solicitation outside the interaction that's more interesting to me but then I will leave the interaction but to maintain the interaction you have to have a capacity of inhibition and executive executive some form of what would be called executive control to be able to listen and and take what the person is saying if you don't have that you're just going to talk think about something else and not let yourself be at take what be constrained by the communicative acts of the other person so back to Dean's point about multi-part and multiplayer communication and engagement when we only have one communication then it can feel like a transmission like the rosetta stone transmitted from the past we don't get a a dot two or a dot one on the rosetta stone structuring learning groups and other activities and relationships and cooperative communication as something unfolding bumping into each other twice says something about those particles that bumping together just once doesn't it's almost like what do they say once is an accident twice is a coincidence three times is a party or something like that so especially with three because one two three there may be a enough of a recurrent engagement and interweaving to support reflection development to de-emphasize any single of those events because none of them are even half of the total series of three so none of them can even be said to be a a plurality or a majority if the dot zero one and two were to vote to vote one of them out or something like that and I think it does point towards the reality of our continued interaction with materials and then towards structures that allow us to engage naturally Rami I think that what communicating in their repetitively with the same people what it achieves it it obviously shapes the the generative model it aligns them in some way so we come to share expectations and share fields of affordances so it's it clearly eases communicative interaction because we don't have to communicate as much when we there there's this trade-off between if you have a lot of shared context you don't have you you you don't have to communicate as much and if you don't have a lot of short context you have to communicate more to uh to to uh to uh yeah have a useful interaction to have a productive interaction and so what I think we we have done is we have like aligned our generative models in ways that helps us understand each other better than we did at the beginning for instance we can refer to things that we said in the past we can say oh when you said that I so I think that's relevant now and so we can make we can point to relations to things that that we said earlier so there's this there there's this continual easing of communication when you have repetitive communicative interaction with somebody that's that I think is why we can come to believe that people we're living with we can reach read each other's mind without without talking because there's so much shared context that something can happen and it can make us think some something that is shared without us even saying anything so there's this this context come to be really complex and really really deep and so yeah interacting repetitively obviously eases communication okay so we can meet multiple times do preparation engagement and reflection before during and after a communication and structure communication so that there are multiple such kind of peaks and troughs we can talk about aligning or synchronizing in the general synchrony sense not making identical but rather coordinating our generative models through perception and learning as you mentioned it helps us understand each other better like a goal of a communication could be epistemic let's get to know each other and it could also have an element of pragmatic value like let's take epistemic action to find out if a collaboration is relevant here and then we're going to engage in these pragmatic actions ramy yeah and i think that a lot of interaction that we share in our like academic and highly linguistic and advanced societies really often look like transmission interaction so we're saying things that we're thinking to each other but i think that these epistemic or communicative in terms of transmission these communicative transmission interactions are the way they are because they are explicit aims of the interaction in which we're engaged that is we're transmitting content now when we're talking to each other because that in a way that that's the goal of the interaction but in in other contexts like it's because we have communicative practices that we have developed that we can have these these goals that can drive an interaction if we're at the gross historic if we're in other contexts we're doing something else but when we're in academic contexts often the goal of the interactions is talking about ideas say transmitting some some some contents transmitting some some some theories for instance explaining theories explaining arguments and so these interaction we we look at them and we're like oh that's clearly there's there we're transmitting content when we're talking but the idea is that the this is really a small subset of the cases of communicative interaction that we engage in in everyday life and we take these as paradigmatic cases of communicative interaction when they are really uh the the pinnacle of the the the linguistic civilization we're in in a really rarefied uh uh context where like academic discussions are not what we do in everyday settings when we we talk to each other and we we do this it's because it's an explicit aim of our interaction now that we do this but in other cases in many in most cases when we speak to each other that's not the aim the explicit aim is not to transfer some information or talk about theories it's about doing things uh together and now we're doing something together we're talking about active inference but it looks as if it can be understood in in terms of content sharing and content transmission but that's because it's the goal of of our interaction it's not because it's the nature of communication I don't I don't know if I'm being clear yep Dean Yager Henryst well I just I think I don't want to put my philosophers cap on but I will just briefly I think we use language and communication to save time it's an exploitation strategy and so I think when we put a 13 week set of parentheses around a learning moment or opportunity we oftentimes get get the card ahead of the horse we don't we don't see the pragmatics because we're busy transmitting and and what ends up happening is the plan forward that affordance that exploitation or optimization or savings of time jumps ahead of the prep the participate in the permeate now we can say that's a luxury but is if if we don't respect the fact that people maybe don't necessarily learn as well when they're handed the plan as opposed to given an opportunity to build one or form one themselves through preparation participation permeation we could go away from this live stream and each of us could literally now design a better plan as opposed to Remy came to us with a plan no what Remy came to us with was a sketch and then we prepped participated and permeated that now the sophistication of whatever comes out of that in terms of a go forward move from each one of us is orders of magnitude greater because somebody didn't say oh well I've only got 13 weeks I don't have the luxury to be able to be pragmatic about this so there I'll I'll take my hat off again I don't mean to sermonize but I think when we actually feel what it's like to let the plan naturally evolve and become a part of the ecology as opposed to driving the ecology the single source it changes everything and over and over again every time we have one of these three set live streams it reinforces what you could do and again you don't even have to be on the the zoom as long as you give over a amount of the amount of time to see and sense what it's like to prep participate and the and the and I think the most important one of all stick around for the permeate because that's where the real nuggets seem to Daniel's metaphor of icebergs things float to the surface when you give it enough time but I understand there's also a 13 week window and I'm paying thousands of dollars to to sit in your class and come on exploit the daylight side of your communication and I just don't know that people don't realize that they're cutting off their noses despite their their learning so interesting a lot there a permeation letting it pervade our generative model or percolate letting those messages be passed letting that ripple or the perturbation take effect at slower and deeper levels and it is really a challenge when the logistical and transmission aspects of communication are scaffolded and those are like the metadata fields that have to be filled out the boxes that have to be checked and they can feel constraining yet that simply removing them is not a positive solution and so the path of least action for many learning scenarios is to replicate the prior over the decades a professor who might teach the same class in the same way and there's pros and cons to that in fact some of the most enlivened learning communities have like a vibrancy that transcends an almost ironically static learning experience like we all read this thing so it's not that that I think the mode of communication and cooperation and education that we're exploring here means that every single experience is going to be like some kind of full attention transformative depth psychology I think it would feel more normal and natural and just like participating in a cool ecosystem that works with that individual's situation what else can we add or where can we go where is active inference in all of this where do you see active inference not like you know what are your intentions with my child but where is active inference in your further research agenda Remy I think it gives the like all-encompassing framework and the abstractness of the framework allows us to apply it to as we said like communication between cells and communication between like countries or groups of individuals or this multi-scale property of active inference I think is really useful in the in allowing us to describe communicative interaction between these various kinds of free energy minimizing agents I saw there's this that's useful I think that's the yeah the learning mechanisms that are at work in active inference are really useful also to explain how communicative interaction is conducted and how repeated communicative interaction can lead to a form of synchronizing of the generative model as has been explored by Vassil in a paper and in 2020 I think so I think that's really useful there's a lot of things that that I think we can take from the active inference framework and yeah the idea that all actually I think the fundamental intuition is that all all action is free energy minimization that's trivial that that's but I think there's a pragmatist insight in active inference that we can use to support the pragmatist view of communication yeah yeah that everything that an organism does it does it to minimize the free energy and active inference is no different and communicative active inference as a form of active inference also aims to minimize the free energy of an organism and so when an organism finds itself in the situation where it's there's a lot of free energy there's a lot of error prediction it will try to engage it in its environment intervene its environment in order to bring itself back in in viable in its viable states and so all these ideas and when we apply them to communication I think they lead not necessarily by implication but I think they there's a path to these ideas through the to the the pragmatic view of communication the idea that that communication is another form of it's just a mechanism of regulation of organisms in order to minimize their free energy I think there's this kind of basic biological understanding of the the behavior of organism I think it leads us more toward the pragmatist view of communication rather than toward the the more informational or transmission view that is that is more at at home in our like really highly advanced and complex linguistic interaction and I think that this like basic biological framework is is a good place to start to understand communication in the pragmatist view very nice answer I'll I'll read a comment from Stephen and then we can give a thought on it as well as provide some closing thoughts so Stephen wrote explicit aims of interaction are important and can be invisible thus communication appears to be based on transmission should we identify the context through the affordances or via naming the water like the water we swim in that is is the context and the explicit aims apparent via the affordances or is there a special sauce that flavors the cultural waters in which we are swimming yes Dean special sauce on that I think what baby did build on what Remy just said I think the perhaps active inference opens up that space to the to the idea of a special sauce it doesn't guarantee that there will be one and it certainly doesn't promise but what it does do is it says what are what are the possibilities within this not just what are we collapsing down to I mean I'm throughout this series I've been talking a lot about but we can't leave the absolutes the things that we've experienced out of this equation but I think what the what I think the active inference affords is that those relativities those and those those entailments and those entanglements always to be determined can be moved towards and and maybe accepted with with more abundance if we use active inference but that's just my optimistic take on it do you have any other closing words before we hear some final thoughts from the others excellent okay blue or Remy with some final thoughts how does this how will it permeate with you okay so I'll go ahead so yeah I was really happy to have the chance to talk about the paper with you I think that one of the most important things that I'll take home is that now I have a really salient solicitation to go see the plant robot in action so like at the second that I close the zoom meeting I'm going to go see that that's really something that I'm taking from the this discussion but yeah no it was really interesting to see different perspectives different backgrounds what what these different different backgrounds can bring to a discussion that is I think that was really enlightening for me and I was really happy to to be there just say thanks for coming Remy we really enjoyed this discussion and for me I think what will permeate is like the you know everything I think about now I just like think about how can I model this in terms of active inference and I feel like this paper will you know influence maybe the way that I think about communication which is a big part of modeling active inference period especially like in a multi-agent dynamical system so I think I will probably leverage this ecological approach in the way that I think about it going forward so I appreciate the opportunity to learn about it awesome my closing thought what I'm left with here is active inference as about affordances it is about the the content of the model is about affordances and active inference is an affordance it allows us to realize epistemic and pragmatic value today and in the future and in that way we can see it as gripper and gripped just like we could talk about the hand as gripper and gripped whether it's holding hands or grabbing an object and especially in this dot two some very nice points added on language as gripper and gripped and contextualizing this kind of breakaway verbal community of research and scholarly education and academia and recognizing that breakaway community and skill as something that's crystallizing out of a much more inclusive and pragmatically oriented niche and I think that will be a win-win framing because it will help ground what can be abstract and lofty discussions and also elevate and bring what bring attention and support to some areas and types of communication that are absolutely natural in the field of affordance view and the ecological view but are not always respected so highly under like a symbolic transmission view and the closing image for me is making the sauce underwater I'm just imagining it's like I don't know SpongeBob SquarePants or someone's trying to cook and make the sauce but we're adding things and it's more just like happening in this diffusive space and in a way all kitchens are in fluids that's why we can smell what we're cooking but it's different enough that the sauce stays in the pan and I think the social inaction of active inference is when the sauce is spilling out of the pan all right Remy you're always welcome back so hopefully you'll share some developments in this line good luck with the rest of your work thanks all thank you all right farewell bye