 Hello and welcome everyone to Active Inference Lab to the guest stream, the Active Inference guest stream number 2.1 on March 18th, 2021 featuring Demetrius Bolas. And this is gonna be a really awesome discussion on a wide variety of topics. So I'll just pass it to Demetrius to introduce himself and take it away. We're gonna have a presentation first and then we'll be documenting questions from the live chat. So please drop any thoughts or questions in there and then we'll have a discussion after this presentation. So thanks again for coming on the stream. We really appreciate the time and please take it away. Hello everybody. I'm very happy being here. Thank you very much the organizers for organizing and presenting such an interesting series of short seminars. So today I would like to present you a work that we have been doing primarily the last six years in Munich in collaboration with Leonard Zilbach and his group in Max Planck. And basically the title I entitled for IEM Human Becoming in and through social interaction is the title of my PhD dissertation that I was presented last year about this month. So for anybody that would like more details at the end of the talk, they would always go back to the document. It's already published in Research Gate and there you can find also the already first papers that came out from this work but more are coming in the way in the next few months. So I will start by presenting the outline of this presentation. So first we will go through the theoretical work which consists of the dialectical at human hypothesis and the dialectical at human hypothesis which both focus and emphasize the importance of social interaction in human becoming across various contexts and various scales from evolution to culture development and psychopathology. Then importantly, we will discuss concrete empirical methodologies with regards to how we can put into the formal test this kind of hypothesis and which we call to personal psychophysiology because this work tries to extend the traditional psychophysiology work in the social domain. So we will see how based on hyperscanning research, for example, we integrate different streams of studying for monitoring, recording and analyzing data and high resolution from multiple people at the same time that they are able to interact in real time. Then we will go through the empirical work and the first two studies which is about the interpersonal at human in real time social interactions, especially through gays. And second, about the interpersonal at human in real life social relationships. And this is with regards to autistic traits and interpersonal similarity. Fourth, we will discuss some scientific and societal implications of this work and we will briefly conclude. So the theoretical work is based on two main pillars. That's a philosophical framework developed across many centuries going back to ancient years and ancient Greece and also ancient China or India. And we will go through more details later, but this was developed further in more modern times, especially in France and Germany. And also we lean upon heavily the Bayesian brain hypothesis and especially the framework of predictive processing and active inference. So I guess most of the people here might be already familiar with these frameworks, but I would like to go through a simple example just to make clear the practical implications of this kind of frameworks. So let us imagine a lady waking up late for her work and so she cannot decide what to wear for going to work, but based on her previous experience and she lives in the south part of Greece in Crete, she believes that most probably the weather will be good and sunny, so she decides to go out without tying an umbrella or a coat. However, this day she's unlucky so contrary to her predictions, the weather was rainy and that day she got wet. The second day, even though she believed that the probability for raining might be 5%, so she increased her probability, she believes about the probability of raining. However, she still decided not to cut the rain and she decided to go out without tying an umbrella however she still decided not to carry an umbrella. This continues for a week where she finally came to believe that the probability of raining, even in this period, in this place, was 90% and decided to carry an umbrella. So we see here how she gradually updated her beliefs based on the evidence that she observed in real life gradually. Let us imagine another illustrative example, the same lady goes for vacations to Japan and she again is late to go out and she has not decided whether she needs to carry an umbrella if she or not. Based on rumors in this time, she again believes that the weather will be good so she's slightly dressed. However, she's again unlucky and the weather is raining. So already, in this case, the lady, for example, from the second day already, she believes that the probability of raining will be 90%. So we see here a jump regarding her updating of the weather beliefs. Why is that happening here? From a Bayesian perspective, one could argue that in this case that her beliefs were formed quickly through social rumors, they were less confident, so more susceptible into change when faced with contradictory evidence. That's a very important factor in the Bayesian brain hypothesis. I will not go through the mathematical details, but with these very simple, simplifying examples, we examine two cardinal aspects of such accounts, namely the updating of beliefs through integrating prior noblets and new data. Also, we examine the learning rate of this procedure, which is thought of as a weighted precision prediction error, which means that the more confident we are about our previous beliefs, the less likely we are to update them in face of contradictory information. So this framework has gained much attention over the last years. First, because as we see, it's useful in examining various real-life scenarios and put them into formal equations and models, which can be formally examined and tested against one another. Of course, the outline of this account that we've drawn so far is rather simplistic. In this case, we only examined the scenario that one person observes passively the world, and their only choice is to update their beliefs according to what they observe. So these skull frisms have mentioned resemblance in some aspects. The old plate of cave allegory were humans, were nothing else than prisoners for the whole lives in a deep cave where they never experienced the outer world, and their only experience about their environment was through projections on a wall, through shadows of the real objects that are reflected on the wall in front of them. So this we can imagine in a way humans or other organisms experiencing the world in a manner like that as we have specific sensory organs where information from the world is projected and we selectively experience the reality around us. Of course, this is just a partial image of what is happening. One right away can think of a second avenue of attunement between the organism and the environment. So as we first said, the organism tries to minimize prediction error through perception and learning by adjusting their own beliefs to the environment. But of course an organism can also update their beliefs through action so that they selectively transform and sample the world in a way that it conforms to their initial beliefs. So this constitutes a more comprehensive account of organismic life and especially human life. But we will argue in line with other recent accounts in this talk that this account is still not enough on its own because it considers individual organisms out of context, out of their social context and especially with regards to humans out of the crucial cultural context as well. In our effort to embed this kind of account we have linked on the electrical approach which as we saw before it constitutes a school of thought which goes back to ancient years and continues until these years always enrich in different historical contexts. So the importance of this framework is that it emphasizes unity of phenomena. It doesn't neglect inner contradictions and emphasize movement of phenomena or change. So a situation of crisis or radical changes doesn't seem to this account as problematic but also as chances of updating and enriching. So this kind of accounts try to overcome the economies such as the organism and the environment, the individual and the collective or as we will see in our work even a patient versus their cultural context. A special emphasis is given not only in intra-level processes but also between level processes across various temporal scales. So this account goes against both dualism and reductionism as it tries to examine phenomena in their wholeness but without metaphysically assuming independent levels of reality. Traditionally, sciences for good reasons also for lack of tools have focused on specific levels of description of reality such as for example the physical or biological level or the psychophysiological level. These kind of studies have provided important insights. However, when examined isolated in single levels we will argue it doesn't provide the whole picture of complex phenomena such as psychosocial cultural or psychiatric phenomena. So in this work we have argued that even deploying simple research paradigms will be crucial to try to probe and monitor processes that live on different levels of description. Based on this assumption we have reviewed human becoming as a dialectic of internalization and externalization and more precisely connected internalization processes to predictive processing which as we saw from our examples is the mechanism of updating our own beliefs to conform to the environmental regularities and we connected the externalization processes and especially collective externalization processes to active inference which is a process of co-regulating the outer world and especially the social world according to our inner beliefs but also according to our collective inner beliefs. One can see human becoming as a human with both physical world and others. So here we see how organisms such as animals or humans try to regulate the world both physical and social world by acting on the world. So for example in this case this little animal holds the ball against gravity first that means that already in an embodied way captures the physical dynamics of Earth, the gravity but also acts against the social inference for example here an attempt to from somebody taking the ball. So we see human becoming as a dialectical human in a nutshell as an interpersonal we say that in the personal statistical regularities say multi-scale hierarchical models on individual level and vice versa and this can happen from across different scales from phylogenesis, culture and ontogenesis to everyday learning, decision making, perception and even so-called automatic interceptive process. Our perspective draws heavily on a Wegochian perspective which has famously declared claimed that through others we become ourselves. So in order to further illuminate this view let us imagine a simple example and let us imagine a little baby that feeling hungry as their interceptive systems is out of balance and tries to reach for food. However the motor repertoire of such a human being it's not developed yet fully and so this baby is unable to reach for food on its own so what happens is that usually the caregiver will notice such kind of behavior and it will try to satisfy that for example by placing the food closer to the baby. So in this case we see we can even imagine how this kind of interpersonal relation between the caregiver and the baby results may result in an internalization on a symbolic level of this kind of process. So the baby both across evolution and development might come to realize that when extending the hands or the index finger calls for attention pointing to a specific object that they need. So we see how an interpersonal relationship based on natural needs for survival might end up in internalization on a symbolic level. So here that example is important in our talk because this illustrates how a collective level of description of process between people might be transformed into single brains and single bodies processes and actually makes us what we are becoming that is humans in our growing up. These kinds of accounts are very much in line with the framework of an activism and especially the account of participatory sensemaking which has focused a lot on the fundamental role of social interactions and both from individual sensemaking which is always affected by coordination dynamics while we are embedded in social interactions or the way to join sensemaking that means how we jointly with others make sense of the world. However this kind of account might seem theoretical. So now here we will ask the question how can we apply this kind of ideas and why is it even important to reflect on this process and whether this process lies on lower level of descriptions or collective levels or between. So to illustrate an example we will use autism as a paradigm example and we will try to illustrate how these ideas can be concretely put into practice. So autism spectrum condition has been described as a pervasive developmental condition which is characterized by difficulties in social interaction and communication as well as restricted interests and all sensory behaviors. So that's the traditional typical definition of autism spectrum conditions and some predominant approaches so far have focused on difficulties or so-called impairments of theory of mind. These theories describe autism as a condition that individuals have problems with understanding the intentions or beliefs of other people. So there are also alternative hypotheses such as the social motivation hypothesis which claims that people with autism have usually diminished social motivation to interact with others. So in this case they might lack opportunities in gaining more cognitive abilities. We see also weak central coherence account and or executive dysfunction which refers to more individual characteristics of autistic individuals such as difficulties with planning or paying more attention to local information rather than global information. We also see broken mirror neuron hypothesis which emphasizes the difficulty of autistic people when it comes to empathizing. More recently we have computational approaches such as the aberrant or inflexible precision updating which grounded in predictive processing accounts they claim that autism might result from an aberrant way of updating precision and more concretely regulating the updating of beliefs as we saw by integrating previously gained experience in face of newly income information. So we see here that this type of account there are either more social accounts like the theory of mind social motivation broken mirror neuron account or even the aberrant precision updating when it comes to the social world. However, even in this case where others are into the frame the condition is exclusively attributed to the environment within single individuals and especially single brains. So the most predominant approach is autism, focus on individuals and try to diagnose or even treat such conditions by paying attention primarily on the individual. So for example in research it's common to identify two groups for example one so called neurotypical or non-autistic group and one autistic group and then by deploying a specific paradigm we try to identify traditionally differences in behavioral patterns or brain function and this kind of paradigms then attribute this kind of differences in a deficient functioning of the autistic. So however as we will argue here and in line with an interactive and subjective term in research of psychopathology and specifically autism we will see how by including real time social interactions and social relationships we can expand this focus from individual brains to the process between people and especially in communication. So this line of work would suggest that it's not enough to identify two different groups of people like neurotypical and autistic people but already by beginning examining diets of people or groups of people one can naturally see that we have available here at least three different broad categories of partners. For example in our case we have neurotypical partners we have autistic partners but we have also mixed partners which means consisting of at least one autistic individual and at least one neurotypical individual. So in this case our hypothesis has been that at least in certain levels especially in the experiential level the autistic individuals will entertain similar levels of smoothness, interactive smoothness as the neurotypical partners whereas mixed diets might appear in certain levels interacting less efficiently. And if that hypothesis is correct then one can see that the deficiency doesn't lie in the individual brain of the autistic but at least partially we can locate it in processes between individuals. This is very much in line with the double empathy problem of Damian Milton who have argued that the ontology of exactly of autism condition does not lie in the individual but it comes from mismatches of social abilities of expectations. One can imagine even probing this framework further by not necessarily dehotomizing people in autistic or non-autistic or any other category we can imagine but rather examining such traits on a continuum and then examining the interaction smoothness based on the distance of people in a space of condition rather than in a single dimensional spectrum of conditions. Later we will see how this can be operationalized. So after this study of example as we already mentioned the last decade especially but also before through more sociological accounts we have seen an intersubjective turn in the study of autism and other conditions just to give a few examples for example the double empathy problem has argued that social difficulties in autism arises to a complication and does not reside in single brains but also as we saw an activist and participatory sense making accounts have claimed that atypicalities can be viewed as a different way of making sense of others and interacting with the world or second-person neuropsychiatry perspectives have claimed that social difficulties are more likely or even only manifest themselves in real-time social interactions whereas more passive scenarios might be less problematic clinically. Finally, part of this work the Electronomy of Human Hypothesis has been together in the subjective approaches and predictive processing approaches have reviewed conditions such as autism but also other conditions not as a disorder of the individual but rather as a cumulative in the person or misat human. So in a nutshell in the person or misat human can be thought of as disturbances of the dynamic and reciprocal unfolding of an interaction across multiple timescales resulting in increasingly divergent prediction and interaction styles or patterns and vice versa. As we saw the hypothesis aligned with other accounts holds clear predictions. So before we start experimenting empirically we made concrete hypotheses in terms of interactions which we suggested that interaction within homogenous and non-tuned diets are expected to or tuned diets are expected to appear smoother compared to heterogeneous diets or non-tuned diets. We will see later what we mean by socially tuned and homogenous or heterogeneous. Importantly we see here an interplay between levels of description between the subjective community gap and individual prediction and interaction styles which might appear diverse across people of different conditions. And here we see how initially the subjective community gap can result in increasing individual diverse prediction and interaction styles and how as people that have formed diverse social expectations about each other how this can result in increasing such an inter-subjective community gap across various scales. One can imagine for example at the scale of a dialogue for example let's say I start with making a comment that another person might take insulting then the other person replies to me in a manner I'm sorry I'm starting feeling angry and this can escalate potentially in a mage of misdemeanour in the scale of just a few seconds or a few minutes. Especially this can happen between people for example that have cultivated distinct systems of beliefs for some different cultures and they don't have the chance to get to know other systems of beliefs. This in the case of autism for example this one can imagine that this kind of interactions might play out during across development so a child in a school for example that other children are not informed about different communication styles might isolate exclude and isolate this child and forcing this child to develop with decreased level of social interaction. We see this not only in terms of autism also about with regards to other social communicative differences people can be excluded for example immigration populations in new places might be excluded initially based on this kind of intersubjective community So in our case going back to autism we will try to personalize the homogeneity of diets of partners as a similarity in terms of autistic traits which we can measure with standard questionnaire tests and we will operationalize in our experiments the smoothness of an interaction based on the perceived friendship quality again through questionnaires for example but later we will see how we can extend this work in a more fine-grained time scales for example in at the scale of milliseconds when it comes to interpersonal gaze behavior or interpersonal facial expression behavior So our experimental methodology takes the form of two-person psychophysiology or collective psychophysiology which tries to expand the physiological phenomena into the sociable collective domain Here our first study on the interpersonal human in real-time social interactions make use of this methodology which is grounded in second person of neuroscience accounts which claim that in real-time social interaction brain or other processes might be fundamentally different than simpler passive social passive observational scenarios So when we really interact with another in real-time for example in the discussion these accounts hypothesize that our brain will function in a different qualitatively different and quantitatively different way as when we just sit back and observe others without the opportunity to feed back to each other I will not go in detail here about the brain process As we in our study here we'll see mainly behavioral results So also our two-person psychophysiology framework is based on previous work of the last couple of decades such as for example fMRI or EEG hyper-scanning Here you see one of the seminal works of Guillaume Dumas which connected two EEG systems into the same computer in order to at the same time examine the brain function of two people while interacting through hand movement Also other later works such as work from Vicky Leong or Takahiko Koikien have examined phenomena for example from deeper in the brain with the use of fMRI hyper-scanning or across development Our two-person setup looks like that So I'll not go here also in detail I just want to make clear that our system is several different recording modalities and in a completely symmetrical way in order to make sure that in preferred synchronicity we can monitor the individual process of two or more people from decision making to eye movements facial expressions and brain function all at the same time so that we can interrelate first processes of different modalities within the same person how decision making is connected to our emotional processing or facial expressions and how we our brain function modulates eye movements and back how our perceived social information modulates brain function and second to interrelate the multi-model process of individuals with one another so how my visual information for example influence your behaviour and how your behaviour influence me in a closed loop so in reality our setup looks like that so two persons are invited in the lab and they solve simple visual tasks while they are able to communicate and interact through a micro-camera at the centre of the screen which deploys the face of the other person in so here in our first experiment the important manipulation was that while people were able to communicate in real times like in Skype manner in half of the game people could not communicate in real time because we replayed the part of the video of the previous interaction without letting people know so in our game the two people thought that they always communicated in real time however contrary to their perception in half of the game they really communicated but in half of the game they just observed the video of the other person so in this case we will be able to disentangle quantitatively the impact of the social interaction dynamics even in the absence of conscious interpretation of the situation so here we can see our two main conditions of the live video feed where it allows the two people to form essentially one system interacting system and in the other case distinct systems, cognitive systems we just observe recorded videos of the other without knowing our game looks like that it's based on a seminal work of Bahador Bahramin and colleagues in London it's a psychophysiological task where we ask people to identify the odd object on the screen without knowing that there is no odd object on the screen so because we have four possibilities left, right, up, down we expect just 25% chance of co-occurrence between the decisions of the two persons so whatever weeks observe above and beyond 25% in a statistical significant manner we can attribute just on the dynamics of the social interaction because every other piece of information remains the same for the two people except for the dynamics of their interaction through games so for example here you can see case one we have covered the eyes of people for privacy reasons but here you can see how people in real time even when they don't see the face of the other person they are synchronized and happy interacting together at the same time look here for example how perfectly aligned is even their facial expressions the same here we cannot see the eyes but the same will see above the eyes the same happens people tend to follow each other eyes but let's see now a second case where people in the same game the same day after a few seconds they played a version of the game where without knowing one person see only the video of the eyes of the other person so in this case you see people start being confused like why don't you follow me as I follow you etc so this creates a mis-at-human an artificial mis-at-human where people naïve about our experimental control cannot attribute to our manipulations but as we saw through questionnaires they attribute through psychological interpretations like the other person was bored or tired at some point of the game here for example we can see how the interpersonal at-human play out in the behavioural domain and specially here in how the two gays traces of the two persons coincide and follow each other like independent interacting little creatures and here are our formal results based on our analysis of the gays trace here we can see we form three experimental conditions we have one pseudo-condition where we suffered diet from different days and we see that the gays coordination the green is the lowest then we see that when people really observe the other person in the room the yellow sorry the orange distribution we see that we have an increased gays coordination metric but what really made the difference is when people are allowed to really interact with each other the gays trace was synchronised the most very interestingly we observe exactly the same pattern in terms of a completely different level of description and this is the decision making alignment again here we see that when we analyse data from people of different partners we see that indeed the probability of agreeing on the same object was around 25% we see here the green distribution centres around 25% then we see that the orange distribution already about the observational scenario is slightly increased but what was highly increased was the decision making alignment in the real-time social interaction scenario also this is reflected in the interrelations between the two modalities we see here both in the online and offline conditions we see a statistical interrelationship between the two modalities so the more the higher the gays coordination the higher was the decision making alignment and we also see that in the case of real-time social interaction this interrelationship appeared stronger also another interesting finding is that the confidence difference between the two conditions was significant so people without even knowing that there were two different conditions appeared more confident about their own choices depending in the real-time social interaction situations as opposed to the real-time or social observation scenarios so here we see how the social interaction dynamics even when we are not aware of them consciously inferences as phenomenologically even to the level of our confidence about our own decision making so finally we also see a relationship between gays coordination and the level of collaboration so the more people follow each other gays the more likely we are to report a higher level of satisfaction in collaboration so from going through the studies too where we examine the personal misadhumance at this time in real-life social relationships here we try to our empirical studies outside of the lab in the real world without performing any manipulation just by collecting registering the experiences the real-life experiences of people in the real world so here we study the link more concretely between the similarity of autistic traits and friendship ratings along a continuum and for that to this end we invited first of friends to complete questionnaires about their individual autistic traits and the quality of their friendship without being able to communicate their replies of course to each other then we analysed this data that we gathered from 144 neurotypical adults which formed 72 same-sex diets of friends whose friendship duration varied a lot from 3 months to 30 years and their meeting frequency again was diverse from a few times per year to daily in the ages of other who from 19 years old to 50 years old and in a balanced in terms of sex sample of participants here we see the distributions of the main measures of our study friendship quality, friendship duration autistic traits and autistic traits mismatch which is simply the difference of my autistic traits versus your autistic traits in a given diet of friendship this study was recently published so people that are interested can go back and see all the details so in a nutshell what we find here is that whereas individual autistic traits were related with the friendship quality in line with previous studies in our study of real-life social relationships this apparent relationship was not significant statistically at least in the amount of our sample size as we see and it only appeared the same very weak interrelationship what was more interesting was the as hypothesized the interrelationship between the mismatch of autistic traits and the friendship quality and here we see that the more the mismatch of autistic traits the more different the two people were in a given diet the less satisfied it appeared in their friendship and this result was irrespective of friendship duration age, sex or even the average dyadic autistic traits similar observations we found with regard to friendship duration there is no interrelationship at all between friendship duration and individual autistic traits in our sample but we indeed found an interrelationship between mismatch of autistic traits and friendship duration so the key constituents of friendship quality that contributed to this kind of results were interpersonal acceptance, closeness and help so based on our empirical observations we can speculate and suggest the future directions of research relevant work which will examine the dialectical phenomena between different levels of analysis from the collective to the individual and back so for example in this case we can speculate that potential higher interpersonal at human between people might lead to smoother social interactions which in turn might transform our interaction with another person in a more rewarding form so we are more happy to communicate with a specific person and this have been found to enhance even interpersonal attraction about each other and this attraction might increase the motivation to interact further and more frequent interaction with the other might boost interaction of success and this interaction success might contribute back to higher interpersonal attachment of course such a process is a simplification and more experiments are needed to examine the different stages of such a dialectical procedure in real life so in our discussion we will examine some limitations and future direction of our work first we can and should measure more fine-grained mechanisms in a more clearly modulatory experiments using hyper-scanning and computational modeling in order to tap into more detailed mechanisms of this kind of phenomena also in the future we can deploy unmediated face-to-face interactions such as real-life monitoring of verbal and touching behavior we should examine also more heterogeneous samples of the population for example mixed gender population or cross-cultural diets and groups or even groups of the same culture but of different socio-economic background or educational background further scales of social interaction dynamics can be incorporated and here we are drawing on the dialectical human hypothesis we suggest to continue with longitudinal tracking of two-person and collective psychophysiology so that we also monitor the development of such kind of a human or misathemen across several years in real-life scenarios also clinical studies are needed in order to examine this kind of hypothesis across as we saw different types of diets or groups which would include autistic or other condition diets typical ones but also mixed ones indeed we have already studied in collaboration with Vicky Leong and Takahiko Koike to deploy such hyper-scanning studies in order to tap into the brain mechanisms using both EEG, hyper-scanning and fMRI hyper-scanning hopefully shown in Japan when the situation stabilizes in terms of the pandemic also we at the same time analyze facial expressions between two people here we already saw how two people might synchronize in facial expressions when they interact in real-time hypothesis are going in a similar line as with what we saw before so before concluding I would like also to show a sketch of our future commutational plans here we have on a paper which is called Beyond 1 by Asian Brain we have suggested that we should go beyond individual by Asian modeling of the isolated brain towards collective models which will capture both intrasubjective beliefs and intrasubjective confidence of beliefs but also intrasubjective coupling factors here in this paper from 2007 we were based on previous work of Donizio and Chris Mattis which is developed around a metabase and two-level commutational scheme for concluding if we should mention that especially in the last two-three years more and more evidence appear in favor of a viewing psychopathology as an interpersonal misathioment and not only as exclusively disorder of the brain and if this evidence at least in certain levels of description holds valid and robust one can imagine that apart from examining treating individuals and diagnosing individuals we can pass through more comprehensive social approaches where we will at the same time treating an individual with a certain condition but at the same time we will also inform the society about these conditions so in a way our approach points towards not tuning only persons with a specific condition into the incidental norm of our current cultural context but tries to foster an interpersonal meeting at the middle so for example of course we don't suggest by deploying for example psychopharmacology in the general population but for example with informational campaigns in schools we indeed suggest tuning of social expectations or enrichment if you will in order to provide a more inclusive society for everybody also our approach points toward treatment of interaction as such in line with for example participatory since making suggestions where we envision modern schools for example that would care about tuning the medium of communication between people for example in a previous work we have suggested developing educational systems mediated through computers or even social robotics or serious gaming between two individuals which could alleviate some initial social pressures and allow people to communicate smoother in the first steps of development of communication between two people as we have seen before in this scheme we can simply observe our suggestion is trying to treat this kind of interpersonal misadument between two individuals one autistic and one non-autistic for example which might diverge if left unnoticed by treating not only the autistic individual but also the neurotypical individual and also trying to take care of the interaction of tuning between them but in a way that at the end we result we are left with a tuned diet and not tuned individuals so what we care is that people can communicate smoothly and satisfactorily with each other and we don't care about identifying ideal normal against which other types styles of communication should conform so in this work concluding taking together we have argued that social interaction does indeed matter across various contexts and timescales and in fact we even consider the question of social interaction as the fundamental philosophical question and our argument here is a generally existential one as without social interaction we cannot even exist as humans. Thank you very much for your kind attention and I would like also to thank several people that have collaborated with us across the last 5-6 years in order to for us to accomplish a vast amount of experiments of about 400 people and especially my academic advisor Leonor Zilbach who assisted me in the journey of the last 6 years thank you very much I would be happy now to receive any question and happily discuss. Maybe you can unshare your screen and we can take it from there well thanks for this amazing talk that was truly something special and we got a bunch of questions maybe you can unshare and then you'll return to regular mode you can unshare your screen in the jits so for everybody who's watching live we have a bunch of questions already and you can ask more and I'll just ask them just in the order that they were presented so Alex wrote a question what are the connections or disconnections of the dialectical perspectives that you brought up and monism or Markovian monism so can you please review this question? yeah how do we think about the similarities and dissimilarities of the dialectical perspectives that you addressed and Markovian monism like kind of Markov blanket everything yes so I have recently come across a couple of works claiming Markovian monism I think from Vanya Vize and colleagues but unfortunately to be honest I haven't had the opportunity yet to go through this kind of accounts in order to draw the similarities but I can tell you for example our approach is a lot of commonalities for example with work of Maxwell Ramstad for example and colleagues who have argued in favor of a multi-scale nested Markov blankets processes so our account shares commonalities with such kind of multi-scale approaches I'm not sure if work also goes through this kind of directions perhaps one one difference might be that in our case we don't take the predictive processing active inference accounts strictly, autologically but we use it as a tool in order to model capture specific parts of the process so we don't necessarily presuppose nested Markov blankets in reality but we still view them as a very useful way of integrating this kind of process formally. Nice that really maps to the realism instrumentalism discussion we've been having in these discussions so here's another question from Sasha what activities or mental states are amenable to creating interpersonal attunement for adults what kind of activities what kind of activities what kind of improvisational or synchronous or you're studying it in a very specific case but as far as just our day-to-day life what kind of activities do you think create that sort of interpersonal attunement for adults so as we saw in the presentation I gave a simple example of interpersonal at humans between a caregiver and a child in order to stress how this kind of interpersonal at human is crucial not just about our interpersonal at human but also about survival so we are we cannot escape actually interpersonal at human because otherwise we don't exist neither biologically but not even as humans so whatever we learn from language and symbolic functions comes from developmentally acquired knowledge and which has been of course being possible through largest larger evolutionary scales but after we gather this kind of capacities I think here actually another important point is that I feel that certain paradigms in the past have been helpful but also misconstitute certain human functioning as starting taking for some processes theory of mind as given statically so and then try to delegate how this kind of theory of mind process play out in interactions so we see here that if we take a more historical developmental approach we see that of course one feedbacks to the other the individual level to the collective but we see here a primacy in a big way primacy of the collective level so first we are in tune with others and then we built ourselves and in the adult life I believe social interaction itself and the experience of the other is what can allow us to attune to others in different many different levels for example as I said before in isolated places that people didn't have the chance to communicate with others might if one wants to use Bayesian terminology they might develop so strict priors in terms of their social expectations but when they face other people from other places they are unable to tune their beliefs one like here just freely speaking for example in this case of so strictly formed beliefs people might face such a tremendous difficulty in updating like they need to use so much energy to update their prior beliefs but they formed across development which might escape to other malicious strategies in order to conform the inner expectations to the outer world even by for example violence or even killing, exterminating the other because they cannot just wrap their minds around the existence of other the manifestation of other types of belief systems for example as we can see in the history for example a lot of times when people from different belief systems or or even different economic interests class together and we see that many times instead of being able to adapt to each other people were violent to each other so I think the here is early age exposure to various forms of life and beliefs so that naturally and organically people can attune to each other. I think as with other processes like language there might even be maybe more liberal but still a window of this kind of for at human so I think we are more open for attuning with others in the world earlier age than later but still through experience both through educating ourselves for example but more crucially such an approach would even go further suggesting that individual training is not enough even if we try to educate ourselves and realize that we have acquired biases against other people this might not be enough even to eliminate all our implicit biases and perhaps only in real practice, in the practice of real time social interaction these models if you want can be rebuilt really important points thanks for sharing that question in the chat from Steven is is there a hypothesis that people on the autistic spectrum exhibit different traits due to perception and action dynamics being different rather than for example just the brain structures limiting their generative model can you please repeat the question one more time is there a hypothesis that people on the autism spectrum exhibit different traits due to their perception and action dynamics being different rather than just their brain structures being different can you hear me? I'm sorry but I'm losing connection at certain points perhaps my internet connection is not full if you could reload on your page how can I access that just hit the reload on your browser on the window nice now I'm seeing you better can you please repeat the last question great the last question was is there any hypothesis that people on the autism spectrum have different traits due to their perception and action dynamics being different rather than just brain differences if I understood correctly I disconnected my camera so that maybe I can hear better the sound if I understood correctly whether autistic people might differ not only in terms of brain function but also action and perceptual styles this exactly what our hypothesis is touching upon it argues that autistic people and other so called neurotypical people might differ across different levels of description from brain function to the way that they generate expectations especially about the social world and to the way that they express these expectations especially about others so ground it in individual accounts of predictive processing and active inference we argue here that and in taking predictive processing and active inference as our two main pillars we see here that when you conform to systems that entertain completely different not completely but different styles of generating and expressing social expectations and this might create a misathuman situation where these styles if left spontaneously evolve in a more and more divergent situation so starting from simply checking the precision updating in single brains can reach through predictive processing to generating of social expectations about others and through active inference to the way we generate or express these expectations through actions to others and especially when we take into account the other also we see how the similarity of these ways can contribute in different scenarios so for example lowering the terminology of the dynamical system for example one can see how taking into how considering this one system how this system can evolve very differently when even when a small mismatch is present in the beginning of an interaction thanks and it just sort of cracked me up because we're almost replicating your experiment we're having a mediated dialectic a dialogue and then now you can see me but I can't see your face and then there's even another level of the affordance with our friends and colleagues who can ask us questions and it's like everybody has their own perspective every question is coming to the top of the mind for someone and then it's through that interaction that we're who we are as individuals and as a lab so just really you know cool to hear how you brought complexity science and all these different world knowledge tradition together to also address like very very prescient social issues so another question from the chat Ivan asks how can I be sure that someone else's social beliefs are the same as mine how can I believe how can I be sure that someone's social beliefs are the same as mine okay yeah so for that I guess you can never be sure like this exactly the old Bayesian framework also that's the beauty also I think that we don't necessarily de-hotomize beliefs between for example sure or not sure but we place on our beliefs different levels of different levels of confidence or precision and like in a recent collaboration with some colleagues as Leo Christoph Moore we examine conceptually the problem of trust and there we argue for example that trust and the personal trust is is fundamental in this kind of relationships so based on our belief of predictability of the other and consistency and also that the other would like to treat as well on a belief of belonging and also on shared values so we just accept that we have shared values with some certain people based on these three points we argue that people should necessarily entertain trust about others in social interactions in order to make our to make our trust interaction functioning well we in fact argue with Leo Christoph Moore that this is inscapable practice in real life social interactions because just because we are not able to track and monitor every possible incident around in our world so the next possibility is just to get paranoid and try to control and monitor the behavior of everybody in order to have access to everybody's beliefs and actions but instead of that I think we as humans deeply trust which of course sometimes is we argue that it's always crucial already from ancient societies like from Tomasello work we see how people based on trust were able to hunt more complex in more complex scenarios and provide for themselves the means to survive in ecological crisis but we also argue that trust also can go wrong in certain times and of course in these cases in betrayal of trust the consequence can be dramatic both in individual and social level so this work where Leo Christoph Moore is the first author will publish soon in a book chapter cool really nice point about the trust in the sense making you're giving sentence after sentence I just have to trust you and anything cognitively that demands more than just simply trusting you is just not going to work so here's a question from Laura in the chat Laura writes thank you doctor bolus I would like to know if it would be necessary to standardize the individuals from different cultures composing a dyad to avoid distractors when you evaluate the interaction so how do we think about culture and the dyad how do we control or explore that yes so hypothesis would go beyond like we use autism as a paradigm example but goes beyond to that and how it would accommodate any cultural or difference or difference in beliefs in a simple way like I said for example by deploying dyad structure where we have homogenous dyad structure for example people of the same culture and people from the same big culture and mixed for example groups and see how different process play out and then we can tease apart at least at certain extent which processes play out within specific cultures or what process might emerge out of just out of the communication of between cultures and then we can maybe see how the meeting of cultures can provide new possibilities for example and just culture is just one aspect in my mind also socio-economic background is also of huge importance so people also don't differ only in cultural practices but they differ a lot in socio-economic background and their intentions and motivations so some people try to change certain aspects of their socio-economic needs because they are not satisfied and even in the level that they cannot survive as a group other people try to preserve and maintain certain socio-economic context so I think it's a very diverse like even in our case that we studied the similarity of autistic traits I consider it as oversimplification we found some interrelationship in friendship quality but still that's just one factor there are so many factors play out in interpersonal at human which is good to try to isolate certain aspects but I think at the end of the day we should try to make to see the big picture in real life it's almost like the ancients and their different concepts of friendship different kinds of friendship or different words for friendship and so quality is going to be a multi-dimensional attribute so Tad asks a question one thing that humans seem to excel at is successfully interacting with complete strangers people we have not yet had time to attune to how does your approach explain or address this yes so as I said like interpersonal human in the early development I think and attachment theories are very important here so I think that we are embedded from day one to the social net social cultural net of our world and that's why we are able to communicate with a complete stranger because just simply because we both are quite common knowledge this can be language for example we can still communicate with an animal for example with our dog but less successfully usually done with other humans but also you see how dogs and humans also came closer across evolution because we gain common ground here and we are more able to communicate so imagine other type of animals more wild animals usually we are more difficult to communicate and make sense of these animals but here again trust is very important just we grew up in completely different places that same with completely different people but and we don't know how each other grew up so but just by sharing common values and common ground just by being able to trust others this can make interaction functioning like when for example people of different cultures invaded other places that they see for the first time people so atrocities for example because people didn't trust the people that they met or they have of course other interest but we see that when we meet an unknown person of course both possibilities are open we can be kind and functioning we can be also very vicious I mean as a human species across our history yeah it's interesting how you brought up the incentives and the intentions almost separately because there's situations that could be win-win but then the communication or the narrative breaks down and then there's atrocity or there are cases when actually the game theory is set up adversarily and it also reminds me of our embodied and encultured nature as humans and what we share even when we don't share language like when I've been traveling and not spoken to language I'll see some street food and it's like you can communicate and with your actions that you think it looks like great food or that you want the food and so it's like everyone tastes you know same taste molecules maybe it's a preference culturally whether it's spicy or sweet but then that's still like a communication and an interaction and an attunement that is with money that one person is not familiar with a language and a context so that's really nice maybe I'll ask one more question or is that cool yes yes two more questions first one is where does active inference come into play or where do you see active inference being able to sort of meet this very large philosophical tradition that you're bringing here yeah so as we saw in the beginning of the presentation I see active inference and predictive processing as let's say the electrical interplay of two strategies as suggested from Carl Fishman and other people before as an interplay for making sense of the and acting up all the world in order for us to survive so I see active inference as the the active possibility we have in order to maintain our living form so not just by conforming with the environment but changing the environment and I found this very neat in this what I found very neat in this account is that we can with the same terminology we can account for both perceptual learning process and active action so we see how our perception learning and action becomes just one whole process that's why I see them as a dialectical phenomenon so inseparable and what I tried when I started studying this process some eight nine years ago in Zurich with Klaseno-Stefan this completely transformed the way I see even the world myself but gradually start feeling that just by focusing on individual processes or even both active inference and predictive processing might limit our scope of interpretation so in my page there I tried to bring together different philosophical or scientific frameworks to try to embed this kind of accounts in a more sociocultural real-life context so in this effort for example also ecological and active approaches that I referred are instrumental also so I would like to see in the future a fusion of philosophical approaches that they see the wholeness of the real-life phenomena such as dialectics with more formal ways of interpreting data such as active inference and also more ecological and active approach and complexly approach such as dynamical system in one story and CSIC for example one way I have seen this happening is that simply perhaps in one paper of ours we suggested that bringing together dialectics and then activisms with active inference we can still keep our emphasis on social interactions and the embedness of organisms in social work but also also using powerful computational tools to examine the individual brains which are embedded in this system so I think an extension of this kind of active inference accounts perhaps through works such as Chris Mattis and Zando Nizzo where they have proposed two level metabasian approaches where by deploying this they suggested this kind of framework for studying both learning and decision making in two levels we have suggested that we can reappropriate this kind of models for studying individual brains and collective action in the same model to our suggestions here that these two levels might not even entertain same mechanisms so in the brain we might have a certain way of minimise prediction error and keep consistent form in social interactions between people might have different mechanisms that might have been described more efficiently with an activist account for example or dynamical systems but I think by bringing together a more complete picture of both the sociological processes and the actual mechanisms at the biological level might and how this feedback might become more apparent Wow, you know I was going to ask what the next research steps were that was quite an agenda I mean it's really awesome just on a closing question note what would you like to just close with or where are we going to go from here what are your next research steps or what are you excited to see the community be looking into Yeah, as I just mentioned I would be very excited to see this kind of school of thoughts coming more closely together I think more and more people are working lately on these bringing insights of more social-intrructional approaches and more formal brain approaches together and also philosophy and practice I think it's very important here as we got to always argue that like a fact is data are not never objective facts so as soon as we gather specific data already we imply theory so only the way we deploy I think philosophy is very important here to form new experiments and new models not only for interpreting data and also this is from the research part but also practice is very important so I would like really this kind of approaches to be used in practice in real life for example in psychiatry is one domain where we can re-examine more holistically certain conditions for example in the past we had more sociological more like social-psychiatry examine also the external factors like the impact of society in specific conditions later we had more cognitive model biological genetic models where we try to identify re-define psychiatric conditions based on the genetics or cognitive abilities I would be very happy to see now these two coming together and one informs the other so a more evidence-based holistic sociological approach in concrete concrete applications of social policy such as psychiatry, education and other societal domains very cool so thanks a lot for joining this guest stream too it was really a great discussion and really we're looking forward to seeing what happens with this line of research and hopefully talking to you and colleagues at future times so we'll certainly see you around thank you very much for inviting and thank you very much for your great initiative I think this seminar which are very illuminating so far awesome talk to you later talk to you again bye