Posters

Manuel BALTIERI

Modularity, the separation principle and active inference

Olgierd BOROWIECKI

Combining environmental and neurobiological perspectives on action automatisation. Ecological-enactive and reinforcement learning views.

Denisa BUTNARU

Corporeal capabilities – embodying exoskeletal devices in spinal cord injury and stroke

Jorge Ivan CAMPOS BRAVO

Communication of conventions through evolution using a minimal cognitive model

Hendry CHAME

A neurorobotics study of cognitive and motor compliance in on-line dyadic interaction

Emma Peng CHIEN

Autism and the Emotion-Enriched Enactivism to Social Cognition

Artur CZESZUMSKI

Embodied priming: How movement based priming influences visual attention

Nagarjuna G

Grounding Syntax through Dexterous Halted-Action-Patterns in a Layered Sensory-Motor-Network

Melina GASTELUM VARGAS

Affordances for learning

Ximena GONZÁLEZ-GRANDÓN

The emergence of self awareness through recognition of self: inner ecology and the role of microbiome interactions

Ken-ichi HARA

Embodied Morality: An Ethical Interpretation of Bergson’s Theory of Laughter.

Jing HE

Acting jointly as a group: Singular or Plural?

Shunichi KASAHARA

Sense of agency in a preemptive body action with the body actuation system.

Alberto MARIOLA

Towards Substitutional Reality: exploring the oculomotor markers underlying change detection in naturalistic scenes.

Keerthy MENON

Open questions in quantum physics from the perspective of embodied cognitive science

Laura MOJICA

No Self without Others: An enactive approach to the external normativity of cognition

Gabriel Axel MONTES

Causal Biomimesis: From Hominin Tool-maker to AI & the Crucial Prospective Role of Body-mind Praxis

Yosuke NAKANO

Broader Cognition

Wataru OHATA

Imitative interaction between a human and a robot based on frameworks of predictive coding and active inference

Prajakt PANDE

Learning with External Representations in Science and Mathematics: An Embodied and Extended Cognition Model of the Cognitive Processes

Aditi POPHALE

Observational Learning in Octopus laqueus

Christin PUTHUR

Mathematical analysis of living motion

Jeffrey QUEISSER

Multi-Modal Sequence Prediction and Action Planning by Manipulation of Working Memory

Susana RAMÍREZ-VIZCAYA

Narrative and embodiment: re-shaping the self in addiction

Georgina Montserrat RESÉNDIZ BENHUMEA

Ant Social Interaction: An Enactive Approach

Luca ROCCIOLETTI

The Hard Problem of Content made harder

Ekaterina SANGATI

Embodied Embedded Modeling of Minimal Joint Action

Hillary SWANSON

Stepping out of rhythm: An embodied artifact for engaging in mathematical reasoning and scientific sense-making

Jasper VAN DEN HERIK

Reflective Situated Normativity

Yuki YAMADA

Embodied cognition in ethnic minorities: A field study about space-valence metaphor

Modularity,the separation principle and active inference

The assumption that action and perception can be investigated independently as separate modules is entrenched in theories, models and experimental approaches across cognitive science. This has been a central point of contention between computational and 4E (enactive, embodied, extended and embedded) theories of cognition, with the former embracing the ``classical sandwich", a modular architecture of the mind, and the latter advocating a more holistic view of cognitive functions depending on agent-environment dynamics. In this work, we first focus on the connections between the modularity of action and perception at the core of computational theories of the mind and the separation principle of control theory. This principle provides formal criteria to define a class of modular systems in computational motor control based on the reafference principle, introducing the use of internal models, their underlying assumptions and limitations. As an alternative approach, we will then consider active inference, a recent proposal of a process theory for sensorimotor control derived from variational Bayesian methods and inspired by the equilibrium-point hypothesis and threshold control. We will show with a simple model that it disposes of the more traditional separation principle, and thus of the assumption of modularity, in favour of an alternative framework based on the minimisation of (proprioceptive) prediction errors, closer to 4E theories of cognition.

Combining environmental and neurobiological perspectives on action automatisation. Ecological-enactive and reinforcement learning views.

Enactive approach emphasis the interactions between brain, body and the environment. Recently, an ecological-enactive perspective has been combined with the free energy principle (Bruineberg et al., 2018). In this view, living organism are understood as the brain-body-environment dynamic systems out of which, the phenomenal consciousness emerges (Kirchoff and Kiverstein, 2019). Part of the environment supporting the existence of the system by offering relevant affordances is called the econiche. Actions performed in the econiche does not require cognition, thus can be operationalised as a habitual (automated) action, performance of which is mediated by the sensorimotor parts of the brain (Dolan and Dayan, 2013). Action automatisation is associated with a gradual shift from DMS (dorsomedial striatum, caudate in primates) to DLS (dorsolateral striatum, putamen in primates) (Poldrack and Packard, 2003). Given above, the emergence of the econiche can be also understood as a gradual automatisation of the actions performed in the environment. For this hypothesis to hold, I suggest a gradual distribution of power of the affordances; spanning from new affordances yet to be explored by the system to known signifiers underlying performance of the habitual (automated) action in the known econiche (Norman, 2013). At the extreme, the econiche shall be understood as a “landscape of signifiers” whereas the “landscape of affordances” shall denote the environment filled with possible opportunities for action, but not necessarily relevant for maintaining the system. In this work I argue that such conceptualisation of the ecological-enactive framework is coherent with the computational accounts of action control: model-free and model-based controllers (Dolan and Dayan, 2013). The ecological-enactive account and the research on instrumental action control share the common subject of inquiry, therefore both accounts benefit from mutual comparison in a way argued in this paper.

Bruineberg, J., Kiverstein, J., Rietveld, E. (2018). The anticipating brain is not a scientist: the free-energy principle from an ecological-enactive perspective. Synthese, 195(6), 2417-2444.

Dolan, R. J., Dayan, P. (2013). Goals and habits in the brain. Neuron, 80(2), 312-325.

Kirchhoff, M. D., Kiverstein, J. (2019a). Extended Consciousness and Predictive Processing: A Third Wave View. Routledge.

Norman, D. (2013). The design of everyday things: Revised and expanded edition. Basic books.

Poldrack, R. A., & Packard, M. G. (2003). Competition among multiple memory systems: converging evidence from animal and human brain studies. Neuropsychologia, 41(3), 245-251.

Corporeal capabilities – embodying exoskeletal devices in spinal cord injury and stroke

Conceptual developments based on Maurice Merleau-Ponty's philosophical insights in phenomenology have stressed the centrality of the body for our experiences. Recent technological developments in the medical field such as exoskeletal devices question identity related notions elaborated in the phenomenological paradigm, among which those of ""own body"" or ""body schema"", are particularly important for one’s relation to spatiality. The formerly evoked examples of technologies redefine embodiment for persons with motility impairments in that they offer a temporary possibility to walk again, although most often in a confined space, that of a laboratory or clinic. The aim of this presentation is to show how the use of exoskeletal devices redesigns walking capability in spinal cord injury (SCI) and cerebrovascular accidents (CVA) and hence forms of embodiment. Exoskeletons are markers of a specific corporeal transition: one where the “I cannot” turns into an “I can” (Husserl [1952] 1989, 266), although temporarily. To show this phenomenon empirical material from narrative interviews with persons having SCI and CVA and expert interviews with engineers shall be considered. The aim is to show how the contextual use of exoskeletons in clinical environment changes the status of the person in that s(he) becomes “temporarily abled” and how the clinical space or the laboratory allows for a short time to recapture former capabilities while forging novel forms of embodiment.

Communication of conventions through evolution using a minimal cognitive model

Conventions in communication is a common practice that we use every day for sharing knowledge. Using a minimal cognitive model we are looking for some insights that show how conventions could be created through the transmission of data (behavior) in the evolutionary process. Using the Iizuka et. al. 2011 model and the Cumulative cultural evolution approach (Kirby et. al. 2008) we created a model where a pair of artificial agents interact in a reduced environment and must share different conventions to accomplish a task. Each agent behavior (sensor-motor system) is controlled by an artificial neural network (CTRNN, Beer 1995) whose parameters were selected using an evolutionary method. For the cultural evolution of the pairs we selected one agent to be the mature user of conventions and the other as the new learner of the conventions. The new learner will become the mature user of the conventions in the next generation and we select a new learner from the pool of offspring created by the previous population. The conventions between the pairs should be created and regulated during the sensor-motor and cultural evolution of the pairs. The success of a pair in the task meant that a convention for that specific task is created and shared through generations. Here are two types of evolutive pressure, the physiological evolution (CTRNN) and the social evolution (conventions). Where an agent should be capable of behave and detect the behavior of the other agent and each convention should be learnable enough to pass through generations.

A neurorobotics study of cognitive and motor compliance in on-line dyadic interaction

Embodiment and subjective experience are important aspects to be considered towards the understanding of intentional actions in dyadic physical interaction. In our research, intended actions are considered within the domain of social cognition, so they are viewed as purposeful acts performed with skill and awareness. From the perspective of neurorobotics, we investigate in on-line human-robot interaction experiments the relation between intended action, compliance (in the motor and the cognitive dimensions), and behavior emergence. We model motor compliance as the capacity to be driven by external physical action, which can be conforming or conflicting with the intended action, through the proposal of a control concept in the joint space based on torque feed-back. We propose a variational model inspired by the principles of predictive coding and active inference to study cognitive compliance, which is viewed as the flexibility in modifying the generative process according to information from the ascending process, that is, the capacity to be driven by sensory evidence (i.e. an inference process). Our experiments with the humanoid Torobo portrait interesting perspectives for the study of developmental and social processes based on rudimentary non-verbally mediated dyadic interaction.

Autism and the Emotion-Enriched Enactivism to Social Cognition

In the philosophy of social cognition, the question of why autistic individuals have difficulties understanding other people is a common exemplar used to evaluate theories of social cognition or how we understand other minds. Currently available theories of social cognition, such as theory-theory, the simulation theory, the interactive approach, and the enactive approach, justify themselves with explanations of why autistic individuals have problems understanding other minds. Among these theories, the enactive approach has provided a more comprehensive account of autism than the alternative views by carefully considering not only social traits but also non-social traits of autism. Specifically, the enactive approach proposes to understand autism through the particular way autistic individuals perceive, move, and emote. However, despite that the enactive approach has provided a more comprehensive account of autism, this approach still falls short in explaining certain aspects of autism, especially autistic emotions, which were suggested to cause autistic social difficulties in Asperger’s and Kanner’s first descriptions of autism. To mend this gap, I shall present a new version of enactivism that shifts the explanatory focus from perception and motion to emotions. This emotion-enriched enactivism combines the original enactive approach and the somatic marker theory, a version of the feeling theory of emotions. This combination expands the enactive approach into a more inclusive view that can account for autistic emotions, the potential cause of autistic social difficulties neglected by currently available theories of social cognition, and thus strengthens the enactive approach against alternative views on social cognition.

Embodied priming: How movement based priming influences visual attention

In everyday life, humans use visual attention to navigate in their surroundings. News web pages are an example of an environment that is present in the everyday life of most humans. The content of news web pages is often emotionally charged, and it is known that different emotions can prime different visual behavior while exploring a web page (Kaspar, Gameiro, and König, 2015). While the influence of emotional priming on free-viewing visual behavior was a topic of research in the past, there is little research investigating embodied primes (Stoykov, Corcos, and Madhavan, 2017). The embodied approach of human cognition suggests that concepts are deeply dependent upon and constrained by characteristics of the physical body of an agent, such as the performed body movements (Foglia and Wilson, 2013). Thus, we think that the effect of emotional primes might be facilitated by combining them with bodily arm movements towards or away from positive and negative cues. In this study, we tried to extend previous research on emotional priming by using the embodied approach. We used the joystick-based Approach Avoidance Task (AAT) to influence the emotional states of participants, and subsequently, we presented pictures of news web pages on a computer screen and measured their eye movements. Twenty participants performed the task into two conditions: congruent and incongruent. In both conditions, we presented a picture taken from the International Affective Picture System (IAPS) (Lang, Bradley, and Cuthbert, 1997). In the congruent conditions, they had to avoid the negative cues (by pushing the joystick away) and approach the positive cues (by pulling the joystick towards themselves). In the incongruent condition, the instruction was reversed: push away positive pictures and pull negative ones. After the priming procedure, we presented participants with twelve structured news web pages. We asked participants to explore the web pages freely. Each web page comprised of two main news (picture and matching heading) and neighboring short information such as weather forecasts, ads, and recent sports results. This provided a natural stimulus set up and allowed to present two contrary emotional-laden pictures simultaneously while tracking the eye movements. Our study design differs from previous priming studies in that it instructs participants to interact with the primes actively. Here, we analyze reaction times in the priming part of the task and eye movements during web pages exploration. Our results suggest that participants need less time to pull the joystick in congruent condition than in incongruent conditions. However, this effect is not present for push movements. Moreover, we explore how embodied primes influence eye movements in a free-viewing task. We analyzed the number of fixations, fixation duration, dwell times, and saccade length on positive and negative pictures embedded in the webpages in congruent and incongruent conditions. We found that saccade length and fixation duration significantly differ between positive and negative pictures, but there was no difference between congruent or incongruent conditions. These results suggest that embodied priming does not influence saccade length and fixation duration in a free-viewing task, but valence of the pictures does. Furthermore, we found that the number of fixation and dwell time as well differ significantly between positive and negative pictures. More importantly, we observed that the number of fixation and dwell time differs significantly between congruent and incongruent conditions. However, we did not find a significant interaction between conditions (congruent/incongruent priming) and valence (positive/negative picture). These results suggest that the number of fixation and dwell time are influenced by the valence of the picture and type of embodied priming. Namely, after congruent priming (pull positive and push negative), participants tend to explore images more, indifferently of their emotional valence. Taken together, we investigate the influence of emotional priming on visual attention. In particular, we extend the priming method with the embodied cognition approach and use movement-based priming.

Grounding Syntax through Dexterous Halted-Action-Patterns in a Layered Sensory-Motor-Network

Quite a few embodied mechanisms and cognitive architectures are being developed and discussed (Merleau-Ponty 1945; Bateson 1972; Gibson 1978; Hutchins 1995; Brooks 1991; Karmiloff-Smith 1992; Clark 1997; Thelen and Smith 1996; O’Regan, Noë 2001; Bongard and Pfiefer 2006; Tani, 2015). These non-cognitivist models give valuable insights, by integrating dynamical systems, neural networks, and distributed processing models, about the roots of experience, learning/memory. However, the semiotic aspects of cognition (Deacon 1997; Wheeler 2006; Corballis 2011) and a core feature of languages such as generative syntax (Chomsky 1988), abstract symbolic operations and culture are weakly addressed by them. To address this gap, we propose a model (as grounded speculation) that generates syntax through dexterous Halted-Action-Patterns (HAPs) embedded in a body characterized as layered sensory-motor-network (SMN). In this proposal, we integrate the following design principles into the existing embodied models: (a) the action patterns of the sensory-motor network are layered; (b) it is to be distinguished from the ’layers’ of neural network in the sense that the layering is based on a gradation of haltability of action-patterns at multiple zones; (c) the physical and physiological (autopoietic) action-patterns (beats e.g. heart-beat, motricity (Llinas 2001)) are considered mandatory and non-haltable forming the core-sustainable layers of the SMN, (d) the haltable-action-patterns (HAPs in contrast to Fixed-Action-Patterns) are pivoted on the core-layer to get disengaged, while still being grounded in the network, making cognitively differentiating actions and experiences possible (making accessible what is available), (e) the HAPs are architectural bases of dexterity of multiple action-zones within the SMN situated in an agonistic and antagonistic manner to enable serial syntactic actions like gesturing, naming, nesting, recursion, reflexivity, etc possible. We discuss the implications of this architecture for furthering cognitive semiotics and cognitive robotics.

Affordances for learning

Lack of a theory of perceptual learning in post cognitivists theories has recently been pointed by Lobo (2019). We insert the present work in this observation, by meaning that our current understanding is ill suited to account to phenomena that involve internal changes in the nervous system of the actor, e.g. learning. Although there have been explanations on the side of learning that belong to skills and performance in certain environments (SIF), there’s still missing investigation on the side of the sensorimotor patterns that develop and evolve with learning and their integration altogether. We argue that affordances 2.0 (Chemero 2009) can be used to explain this process. First, we expand on how affordances are temporal and imply nested processes, second, we will take learning to be understood with the key concept of synergies (Corris and Chemero 2019) and the evolution of those particular states of cognition in certain timescales. Third, we will point out that this has a consequence a different phenomenal valence and affect as learning evolves through time, so it might make sense to say that the entire organism-environment system retains changes in its structure as a result of the organism’s learning of sensorimotor contingencies that in turn give rise to a faster and more automatic solicitation of the affordances that are available in the environment.

The emergence of self awareness through recognition of self: inner ecology and the role of microbiome interactions

It is now recognized that all multicellular eukaryotic organisms are holobionts exhibiting complex and emergent interactions between symbionts, the hosts and their microbiome, antigens and their proper cells co-evolving in living ecologies. This conference contribution aims to outline a scientific approach to the investigation of self-awareness, emphasizing the role of self-organizing biology and microbiome-host interactions in identity. We aim to show that for understanding the origin and evolution of self-awareness, we need to consider the interface between the theory of self-organized criticality (SOC) and the network-centered perspective of immune systems. In this sense, we propose to contemplate (a) that the immune system’s activity defines its cognitive domain: the dynamic interaction of immunocytes antigenic environment within the immunological and microbiome context is a network that ensures the identity and integrity of the holobiont organism (“self”). And, starting from the molecular and multicellular organizational level’s proclivity to efficiently maximize the use of all forms of energy, we consider that (b) these ecologies are placed at the phase transitional or critical self-organizing points. It is at these borderlines that competing requirements of robustness and adaptability are optimized, leading to adaptive multicellularity and self-awareness emergency. According to this approach, information and energy transfer in living forms can be best framed as a phase transitional point in the physical order, and as an ecological system in the biological order, about regular relations between aspects of the shared environment. This integrative framework allows us to interpret self-awareness, as a prerequisite of true consciousness, to be identified as a form of collective cellular swarming organization– the critical multitudinous interactions of the proper cells, immunocytes, and microbiota symbionts.

Embodied Morality: An Ethical Interpretation of Bergson’s Theory of Laughter.

What makes morality possible? The question invites a standard reply. Some philosophers would claim that what makes morality possible is the highly intellectual faculty known as reason. I propose that we instead pursue another position. I advocate a very different sort of understanding of our morality: an embodied morality, based on the French philosopher Henri Bergson’s theory of laughter. 　　　My presentation begins by reviewing Bergson’s two concepts of sociality: closed and opened sociality. Both of these concepts entail an understanding of embodied morality. Society is closed automatically in the same way as bodily habits, and it is opened by the non-propositional utterances (of mystics). Next, I suggest that the same scheme can be found in Bergson’s theory of laughter. We reflect on ourselves when we are laughed at. However we cannot find any propositional utterance in laughter (instead, we merely hear the ‘ha ha’ of the other), and our reflection can be performed automatically as soon as we hear laughter or see a face laughing at us. My presentation then proceeds to argue that Bergson’s theory of laughter offers a concrete example of an alternative framework for understanding morality. As I said above, some philosophers have assumed that morality consists in our highly intellectual faculties. In other words, the moral properties have been regarded not as the object of perception or sensation but as that of thought or judgement. On the contrary, from Bergson’s point of view, our morality includes, at least partially, the embodied aspect of our life.

Acting jointly as a group: Singular or Plural?

The social world cannot be adequately studied without making use of the notions of “having an attitude and acting as a group member” versus “having an attitude and acting as an individual”. In this talk I will identify the we-mode which essentially involves the idea of thinking and acting as a group member, thus this mode will be understood to contain at least the ethos of the group, collective commitment, affective elements (e.g. “we-feeling” or “the feeling of togetherness”), and action based on these components. On the contrary, I-mode involves the idea of thinking and acting as a private person. After considering carefully Brandom’s normative pragmatic account of sociality as “I-Thou”, it seem to me that he sets up the only alternatives for sociality either individual interaction (plural) or a superperson (singular). My worry concerns the coherence of Brandom’s proposal is: If “I-Thou” mode is irreducibly collective (singular), then, is it a suspicion that the idea of a collective factor playing important roles is theoretically unsound? If “I-Thou” mode is plural, and all its manifestations and functionality are reduced to individual representations, then how the “I-Thou” functions, gets coordinated, or is even possible? By drawing on the early contributions of the pragmatist Mead, I will intent to argue that social encounters often find us adopting an attitude in which we move in a ‘we-space’ of action, attention, affect, language, and concerns. This is welcome in the context of a move away from individualism in social cognition research. Finally, I will try to clarify, to what extent, we-mode action is more fundamental than I-mode action to social world in the sense that cooperation, collective intentionality and group responsibility require a underlying we-mode.

Sense of agency in a preemptive body action with the body actuation system.

The sense of agency refers to the sensation that I caused the action. Generally, one would expect that if own body is moved passively by an external force, one would not feel a sense of agency. However, Kasahara [1] have reported that even with the EMS-based preemptive and passive body movement approximately 80 ms faster than the human voluntary movement during the falling-ball catch task, the participant still reported a sense of agency. This effect however only has been verified in a simple reaction task in which the participant and the device share the same objective (i.e., both participant and device press a button as fast as possible). In order to investigate how preemptive body actuation impacts the sense of agency in a more complex situation in which a participant cooperates with a body actuation system, we created a finger magnetic exoskeleton system that can actuate fingers involuntarily. In this poster, we introduce work in progress of psychophysical experiments of reported sense of agency with the simple reaction task and the cognitively demanding task where one’s finger is actuated faster or slower than the user's own reaction. We report preliminary results of experiments with our magnetic exoskeleton and comparison with EMS based actuation. Our result suggests the possible designs for computer assistance of body motion with preemptive actuation.

[1] Kasahara et al. Preemptive Action, CHI2019

Towards Substitutional Reality: exploring the oculomotor markers underlying change detection in naturalistic scenes.

Substitutional Reality (SR) is an experimental paradigm built with the goal of manipulating and assessing people’s conviction of reality (CR). The system uses a head mounted display coupled with a frontal camera that allows participants’ visual experience to be continuously switched between live input and pre-recorded videos of the same environment, without participants noticing the change. These manipulations in sensory input can thus be used to dynamically modify CR. Despite these promising capabilities, obtaining an explicit report of participants’ belief state in SR is challenging as this would compromise the experience itself by making participants aware of the introduced changes. To address this issue, we investigated the implicit oculomotor markers of change detection in complex scenes. 20 participants (10 per experiment) completed a change detection task in natural scenes, presented either within a partial SR setup (without “live” video component) or on a flat screen. Participants continuously explored panoramic images of different rooms with their eyes to detect an object substitution (e.g. a small bin substituted by a box) introduced during the trial using gaze-absence contingency. Preliminary results showed that both pupil size and fixation duration towards the target area tracked participants’ detection performance and the confidence in their decision - larger pupil dilation values and longer fixation duration in the post-change period indicated correct responses with maximum confidence. These results provide a promising pathway to investigate decisions and belief updating during real time exploration of natural scenes, giving promise to SR as an experimental method.

Open questions in quantum physics from the perspective of embodied cognitive science

Quantum theory, by far the most successful paradigm in physics, still faces the challenging question of reality. The common-sense notion that the world is independent of an observer, a naïve form of realism, turns out to be in conflict with quantum mechanics. This is known as the measurement problem. But there are also tensions within physics: Einstein locality, which states that no influence of any kind can propagate faster than the speed of light, is contradicted by quantum entanglement. We speculate that these problems could be related to an inadequate understanding of the human observer. This suggests that we should take a closer look at what the latest theories in cognitive science have to say about the human mind. It turns out that during the last couple of decades there has been a paradigm shift: Cognitive science is trying to bridge biological cognition and consciousness, and, importantly, it treats their basis as not just limited within the brain or even body, but rather as extending outwards into the subjects’ environmental interaction. In this poster we will use this new perspective in cognitive science to consider the question of the hour in modern physics: why does the quantum superposition of a state collapse when you are measuring it using a classical apparatus? Can quantum superpositions be scaled up to objects in the classical world? And how, if at all, does the conscious observer become involved in this process?

No Self without Others: An enactive approach to the external normativity of cognition

Despite providing a naturalistic explanation of the emergence of a normative point of view, enactivism has been accused of being excessively individualistic for defining normativity in terms of self-maintenance of the agent’s biological and sensorimotor identity. As noted by Heras-Escribano and colleagues, it has the problematic consequence of leaving out external correctness conditions of cognition: any behavior can be interpreted by the individual as contributing to her self-sustenance, thus everything she does is right if she believes so. Enactivism can address this problem without renouncing to the centrality of identity self-maintenance if sensorimotor identity is understood as essentially social. An agent’s sensorimotor identity is the patterns of activities that make up her forms of life. I argue that these patterns are constituted in embodied and situated practices shared with others. This means, first, that how sensorimotor coordinations organize and which become frequently enacted is determined by embodied and situated intersubjective interactions during development or any learning process, and second, that a given action contributes to the maintenance of an agent’s identity if it allows her to continue belonging to a shared embodied form of life, i.e., actually or potentially interacting with others she cares about. As social practices are embodied and situated, they incorporate the normative constraints that nature can impose on self-maintenance. Thus, a cognitive agent is intrinsically moved to act in accordance with socially shared patterns because it is essential to her identity, but the correctness of her cognitive behavior depends on the interaction with her community.

Causal Biomimesis: From Hominin Tool-maker to AI & the Crucial Prospective Role of Body-mind Praxis

For millions of years, hominins have been engaged in tool-making and concomitant experimentation. This cognitive enterprise has eventually led to the creation of 'artificial intelligence' in the form of complex computing and artificial agents, whose purposes include solving problems and elucidating the workings of human biology and consciousness, with the promise of surpassing human intelligence. However, much of the expensive efforts in AI development generate techno-social effects that outpace (and may preclude) the benefits of advanced technology for human wellbeing. I present and expound upon the ‘causal biomimesis’ hypothesis: with respect to the relationship between humans and artificial intelligence, the virtually inevitable intrinsic evolutionary consequence of tool-making (and therefore AI development)—and their primary promise—is the externalization of human cognition and agency into synthetic form. The mechanism for this post-biological reproduction is the capacity for dualistic objective thought, which is progressively cultivated as an agent's sense of self and tool-making symbiotically fashion one another, and is carried out through the process of anthropogenic niche construction. In light of this radical hypothesis, existential and ethical implications are considered. Of particular emphasis is that, given the bidirectional influence of human and AI cognition on one another and the risk that carries for the former, it is crucial that the full capabilities of human consciousness are unearthed, recognized and cultivated by humankind, researched, and implemented into AI development. The case is made that much of this potential is found in the troves of embodied practice methods held in the world’s wisdom traditions.

Broader Cognition

For the discipline of cognitive science that has been anthropogenic, Lyon (2006) suggests that the investigation to reveal even human individual cognition has to implement a biogenic perspective. However, even though cognitive scientists have been admitting the other organisms, including plants (Calvo, Sahi, & Trewavas, 2017; Garzón & Keijzer, 2009) and single-cell organisms (Adamatzky, Armstrong, Jones, & Gunji, 2013), as cognitive entities, it seems that they have not provided a clear concept of cognition. Thus, whereas other hard sciences can recognize their objects of studies, cognitive science is still struggling to objectify its object of study. In this current presentation, I suggest a plausible definition of cognition while referring to the Brussel School’s idea of the dissipative structure and complexity (I. Prigogine & Stengers, 1985; Ilya Prigogine, 1997), its reasoning, and the prospective contributions. I define cognition as the disposition of a complex system to alter its state into another while its functionality. This idea may provide the explanation of what drives cognitive system to function. Also, each type or scale of systems has peculiar aspects such as components and media of the interaction; however, defining and consider cognition from the more reduced physical level point of view, we may observe and discuss cognitive phenomena in different scales with the same frame of reference.

Imitative interaction between a human and a robot based on frameworks of predictive coding and active inference

This study proposes a visuo-proprioceptive model based on frameworks of predictive coding and active inference by means of a variational Bayes recurrent neural network. We investigated the model in imitative interactions between a human and a humanoid robot in simulation. The model consists of proprioceptive modality, visual modality, and associative part. It dynamically predicts visual perception and proprioception simultaneously through generative processes, and updates the internal model and generates actions by maximizing the evidence lower bound. A key feature of the model is that each modality is capable of controlling its own complexity independently. We argue that controlling the complexity of modalities contributes to two aspects of agents with multi-modalities in social cognition. First, it helps the model coordinate multiple modalities well. Our results suggest that the visual modality should have less complexity than the propriocpetion modality. Second, it plays a significant role in how the agent should change its intention in human-robot interaction scenarios. Based on different degree of complexity, we modeled agents with strong top-down prior, and those with weak prior. In imitative interaction experiments, while the agents with strong intention tended to ignore the human movements and keep realizing their intention, those with weak prior tended to change their intention and follow the human movements. These results suggests that agents under the frameworks of predictive coding and active inference are required to have the ability to control the complexity to perform adequately in human-robot interaction.

Learning with External Representations in Science and Mathematics: An Embodied and Extended Cognition Model of the Cognitive Processes

This presentation takes an embodied and extended cognition perspective to ER integration – a cognitive process through which a learner integrates external representations (ERs) in a domain, with her internal (mental) schema, as she interacts with, uses, understands and transforms between those ERs. In the talk, I argue for a theoretical as well as empirical shift in future investigations of ER integration, by proposing a model of cognitive mechanisms underlying the process, based on recent advances in distributed, extended and embodied cognition. I present this new model in contrast to the still dominant classical cognitivist (information processing) approaches to ER integration, and various educational technology intervention designs such approaches inspire. I then exemplify the contrast between the information processing model and the new model through a case of arithmetic problem solving, in the light of corroborative neuroscience evidence which demonstrates the criticalilty of action/sensorimotor mechanisms to ER integration and learning. Finally, as educational implications of the new model, I demonstrate the need for: (i) re-viewing the development of ER integration and expertise also as a fine-tuning of the cognitive agent’s sensorimotor (action) system, and (ii) a shift of focus in new-media intervention design principles based on the newer understanding of ER integration in science and mathematics.

Observational Learning in Octopus laqueus

The evolution of octopuses from sessile, shelled mollusks to soft bodied animals with potentially infinite degrees of freedom in movement and posture led to changes in how they perceived and had agency in their environment. Their nervous system became decentralized, that is, two-third of their neurons are distributed in their eight arms. Octopuses are known for their unexpectedly complex and variable behavioral repertoire. Behavioral complexity of octopuses is exceptional not only because of their evolutionary distance from other extant intelligent life but, because their life histories do not have the classic hallmarks of cognitively complex animals such as chimpanzees or dolphins. Octopuses, like the other cephalopods, have short life spans during which they are largely solitary and receive no parental care. Therefore, this unique group can shed light onto embodiment of cognition. The first step would be to empirically test the extent of their cognitive abilities. Octopuses have been shown to be capable of associative learning and there has been some indication that Octopus vulgaris can learn to perform certain tasks by watching other octopuses. However, this result has not been adequately reproduced and studied in only one species of octopus. The aim of this study is to investigate whether octopuses can learn to perform a task by watching other conspecifics. This work will provide unique insights into the cognitive capabilities of octopuses and establish them as a model to study more complex forms of embodied learning, imitation and cooperation.

Mathematical analysis of living motion

We are able to identify living systems by sight. The question of whether a computer can do the same is relevant in the field of artificial life. An interesting online competition, the "Fake Life Recognition Contest", sets up the challenge to distinguish between real and artificial 2D trajectories. In our attempt at the challenge, we tried to distinguish between the trajectories by searching for power laws from the data that would imply Lévy-walk characteristics as seen in the foraging behavior of certain living organisms. For each trajectory, we plotted the number of time intervals as a function of the time interval lengths during which there is a minimum change in angle. The trajectories that showed a power law for this plot with a very good or a very bad fit may have been generated artificially. We made the final distinction between real and artificial trajectories by comparing the magnitudes of the value of chi-square for the power law fits.

Multi-Modal Sequence Prediction and Action Planning by Manipulation of Working Memory

Cognitive abilities, like the generation of goal-directed behaviors, benefit from mental simulations that integrate actions and their consequences. The required action plans for self generated or external goals can be estimated by modifying the internal state of the mental model to generate a sequence of actions that most likely attain the goal. Our current study proposes a novel integration of working memories and the manipulation of thereof for an improved generalization capability of mental models. By a separation of the working memory into a static and a dynamic scope and a specific restriction of the connectivity, a manipulation of the working memory emerges. To this aim, a neuronal network model based on variational Bayes predictive coding is utilized that generates sensorimotor signals, modulates the attention, manipulates the working memory, and controls the fusion of the visuomotor pathways. Besides the autonomous top-down projection guided by attention and the dynamic organization of an occlusion-free visual working memory a cognitively meaningful form of variable binding is observed: the system predicts, as far as feasible, manipulation sequences of the content of future states instead of the content itself. We demonstrate that our proposed method leads to a self-organized and spatial distributed representation of actions and identities for object manipulation tasks, and results in an improved generalization of actions to unseen objects in comparison to previous models.

Narrative and embodiment: re-shaping the self in addiction

Most research on addiction is based on a brain disease model that explains it in terms of dysfunctional neural networks. The need for an integrative model has been recognized, but none of the current models satisfactorily integrates the different levels of analysis without being an aggregation of factors or privileging one over the others. An enactive approach to habits has been proposed as an integrative framework that considers addiction from the perspective of the whole brain-body-environment dynamical system. This approach sees habits as autonomous, adaptive networks of neuronal, bodily, and interactive processes that self-organize to conform an ecology of regional identities from which a global identity or self emerges. Addiction is considered a ""bad habit"" that severely restrains the expression of other regional identities relevant for wellbeing, so treatment should aim at a holistic reconfiguration of the agent's self. In this presentation, I argue that narrative therapy is a good candidate to exert such a reshaping. This may seem counterintuitive given the folk-psychological understanding of habits as independent of reflection. Consistent with this view from folk-psychology, habits are mainly seen as sedimented on a procedurally based body memory that belongs to a pre-reflective domain, while narratives are largely seen as decoupled, higher-order cognitive abilities that cannot exert any structuring effect on the embodied self. I propose that an understanding of the relation between different levels of embodiment or identity generation can help us to understand how narratives can transform our habitual body.

Ant Social Interaction: An Enactive Approach

Social interaction has been widely studied in social insects, particularly, ants. Many studies have shown that chemical communication plays a fundamental role in ants’ interaction networks and colony behavior. However, there are still missing pieces that could help to fully grasp the role of social interaction in itself in ant colonies. Our proposal is to study social interaction in a swarm of ants from an enactive approach, by keeping the basics of Candadai et al. (2019) model, in which they proved that interaction increases the complexity of an agent’s neural activity and revealed that this cannot be achieved in isolation. We subsequently focus on different modes of coupling, specifically, pheromonic and saliva-based, in order to demonstrate how do these different modes of coupling influence the complexity that can be generated by their interaction. Furthermore, we aim to test the cases of interacting and isolated ants that might help us to get new insights.

The Hard Problem of Content made harder

Cognitivist theories of mind have been attacked from different sides. In the talk I will focus on the Hard Problem of Content (Hutto and Myin 2013/2017) and on the generality problem put forward by relational theorist of perception (Travis 2004/2013, Brewer 2006). Both kind of objections, I will propose, take aim at the claim that cognition is essentially a matter of processing truth bearing information through representational states. While Hutto & Myin attack representationalism from the perspective of cognitive science, pointing to the failure of the most prominent attempts to provide a naturalistic account of content, the relationalists exploit the categorical difference between what is general and what is particular, holding that a representation of something will always fall under the former, while our perceptual experience pertains to the latter. The Hard Problem of Content is forcing cognitivists to set their priorities straight. To talk representationally about the products of the supposed information processing by the brain, such as our experience, a naturalistic credible notion of content is needed. The generality problem instead is providing an a priori argument for why conceiving of our experience in representational terms doesn’t make sense in the first place. It pushes the cognitivist endeavor one step further back. I will conclude by arguing that through those objections it is possible to see a substantial theoretical common ground between relationalism and enactivism and that, moreover, relationalism can complement enactivism by providing a philosophically robust epistemological conception of experience.

Embodied Embedded Modeling of Minimal Joint Action

Joint action is a form of social behavior in which two or more individuals coordinate their actions to accomplish a task together. Traditionally, this behavior is explained by postulating cognitive representations of the shared environment, the joint goal, the task distribution and so forth. Embodied Cognitive Science (ECS) denies the need for cognitive representations in producing adaptive behavior in general and social adaptive behavior in particular. One way to demonstrate the feasibility of an ECS view is to construct non-representational Minimal Cognitive Agents (MCAs) that successfully perform some “representation-hungry” task and show that even if such agents possess internal states, these are not usefully interpreted as “representations” nor do they fully explain resulting behavior. In this paper we present an MCA model of a minimal yet representation-hungry joint action task. The task environment is a bounded line which contains a target that moves continuously between left and right borders and a tracker whose velocity is controlled in a complementary manner by two agents. It has been suggested that successful tracking requires that the agents “co-represent'' each other's contributions to the joint tracker behavior. We show that pairs of agents implemented as Continuous Time Recurrent Neural Networks with no such built-in co-representation mechanism are able to evolve a solution to the task. However, we also find emergent neural activity patterns that are consistent with it, namely a predictive statistical relationship between the network activity of one agent and the behavior of another. Whether such patterns are truly representational is further discussed.

Stepping out of rhythm: An embodied artifact for engaging in mathematical reasoning and scientific sense-making

Interpreting representations is a cornerstone science activity. However, interpretation is difficult for learners, particularly when representations are mathematically formalized. This difficulty is well-documented, as is its resistance to direct instruction. It has been argued that the representational form of knowledge can make that knowledge more or less cognitively accessible. A shift in representational form that renders knowledge more accessible (e.g., the shift from Roman to Hindu-Arabic numerals) has been called a restructuration. We introduce a novel restructuration that transformed a mathematical representation into an embodied artifact. Embodied artifacts are performative, embodied representations of a concept. Our embodied artifact helped 8th grade students interpret a challenging dataset to find that a liquid’s rate of warming decreased over time. Students “acted out” the temperature change, walking the length of a giant horizontal thermometer while their classmates clapped out the passage of time, and then reflected on the activity. To arrive at the correct temperature at the right time, students had to continually slow their pace, which led to them noticing the change in rate of temperature change (formally: negative acceleration). Our analysis develops an account of students’ transition from action to concept, guided by the changing rhythm in their body. We show how the activity served as a preparation for more formal learning of thermodynamics, by activating intuitions that students leveraged in their construction of a scientific explanation for the equilibration phenomenon. Embodied representations, we argue, open an alternative route for students to engage in rich mathematical reasoning and scientific sense-making.

Reflective Situated Normativity

Situated normativity is the ability of skilled individuals to distinguish better from worse in the context of a particular situation (Rietveld 2008). Situated normativity is expressed in situated appreciations. A situated appreciation can be experienced as a bodily affective tension that motivates a skilled individual to act on particular possibilities for action, called affordances, offered by a concrete socio-material situation (Rietveld & Kiverstein 2014). Which affordances motivate an individual is determined by their embodied skills and habits that have been shaped in a history of learning in socio-material practices. In earlier work, the notion of situated normativity was primarily discussed in the context of unreflective action. In this paper, we aim to explore and sketch the role of the concept of situated normativity in characterising reflective forms of normativity. In a first step, building on the expressive view of language (Taylor 2016; Merleau-Ponty 1945), we show how language enables us to jointly articulate a situated normative appreciation in a concrete situation. By means of concrete examples we show how appreciations often emerge from interaction with others and the socio-materiality of the particular situation. In a second step, we turn attention to explicit rules. In contradistinction to the idea that reflection marks a different kind of cognition from absorbed coping (e.g. Dreyfus 2002), we argue that formulating and enforcing explicit rules are situated and skilfully performed activities that are afforded by sociomaterial practices (cf. Noë 2015). This in turn enables us to characterise reflective forms of situated normativity as continuous with unreflective forms of situated normativity.

Dreyfus, H.L. (2002). Intelligence without representation – Merleau-Ponty’s critique of mental representation. Phenomenology and the Cognitive Sciences, 1, 367–383. doi.org/10.1023/A:1021351606209

Merleau-Ponty, M. (1945/2012). Phenomenology of Perception (Trans: D.A. Landes). Routledge.

Noë, A. (2015). Strange Tools: Art and Human Nature. Hill Wang.

Rietveld, E. (2008). Situated Normativity: The Normative Aspect of Embodied Cognition in Unreflective Action. Mind, 117(468), 973–1004.doi.org/10.1093/mind/fzn050

Rietveld, E. Kiverstein, J. (2014). A rich landscape of affordances. Ecological Psychology, 26(4), 325–352. doi.org/10.1080/10407413.2014.958035

Taylor, C. (2016). The Language Animal. The Full Shape of the Human Linguistic Capacity. The Belknap Press of Harvard University Press.

Embodied cognition in ethnic minorities: A field study about space-valence metaphor

Upside and downside are associated with positive and negative emotions, respectively (i.e., space-valence metaphor). However, some people are likely to have the opposite association because of their historical background and living environments: For example, Hani, Dai, Yao, and Zhuang people in China. Here, the present study examined whether these ethnic minorities would hold the typical space-valence metaphor. As a result, Hani, Dai, and Zhuang people had the typical space-valence metaphor, while this metaphor was not found in Yao people. Based on our findings, we will discuss how historical backgrounds and living environments are involved with space-valence metaphor.