Beyond Autism: Introducing the Dialectical Misattunement Hypothesis and a Bayesian Account of Intersubjectivity

Review Psychopathology 2017;50:355–372 Received: September 22, 2016 Accepted after revision: October 17, 2017 DOI: 10.1159/000484353 Published online: December 13, 2017 Beyond Autism: Introducing the Dialectical Misattunement Hypothesis and a Bayesian Account of Intersubjectivity Dimitris Bolis a–c Joshua Balsters c Nicole Wenderoth c Cristina Becchio d, e Leonhard Schilbach a, b, f, g a Independent Max Planck Research Group for Social Neuroscience, Max Planck Institute of Psychiatry, and b International Max Planck Research School for Translational Psychiatry (IMPRS-TP), Munich, Germany; c Neural Control of Movement Lab, Department of Health Sciences and Technology, ETH Zurich, Zurich, Switzerland; d Department of Psychology, University of Turin, Turin, and e Robotics, Brain and Cognitive Sciences, Fondazione Istituto Italiano di Tecnologia, Genoa, Italy; f Department of Psychiatry, Ludwig Maximiliam University, and g Outpatient and Day Clinic for Disorders of Social Interaction, Max Planck Institute of Psychiatry, Munich, Germany Keywords also as a dynamic interpersonal mismatch that encompasses Autism · Dialectical misattunement · Social interaction · various levels of description. Moving from a mere compari- Intersubjectivity · Cultural historical activity theory · son of groups, i.e., “healthy” persons versus “patients,” to a Enactivism · Predictive processing/coding · Active inference · fine-grained analysis of social interactions within dyads and Dialectics · Vygotsky · Bayes groups of individuals will open new avenues and may allow to avoid an overly neurocentric scope in psychiatric research as well as help to reduce social exclusion. Abstract © 2017 S. Karger AG, Basel Drawing on sociocultural theories and Bayesian accounts of brain function, in this article we construe psychiatric condi- δὶς ἐς τὸν αὐτὸν ποταμὸν οὐκ ἂν ἐμβαίης. You could not step twice into the same river. tions as disorders of social interaction to fully account for τὰ ὄντα ἰέναι τε πάντα καὶ μένειν οὐδέν. their complexity and dynamicity across levels of description All flows, nothing stays. and temporal scales. After an introduction of the theoretical Heraclitus (ca. 535–475 BC) underpinnings of our integrative approach, we take autism Through others, we become ourselves. spectrum conditions (ASC) as a paradigm example and dis- Lev Vygotsky (1896–1936) cuss how neurocognitive hypotheses can be translated into a Bayesian formulation, i.e., in terms of predictive processing and active inference. We then argue that consideration of A Synthesis of Dialectical and Computational individuals (even within a Bayesian framework) will not be Perspectives enough for a comprehensive understanding of psychiatric conditions and consequently put forward the dialectical Psychiatry through a Dialectical Lens misattunement hypothesis, which views psychopathology In this paper, we will put forward an integrative ap- not merely as disordered function within single brains but proach for revisiting psychiatric conditions, taking dia- 193.175.245.253 - 1/11/2018 5:02:03 PM © 2017 S. Karger AG, Basel Dimitris Bolis Max Planck Institute of Psychiatry Kraepelinstrasse 2–10 E-Mail karger@karger.com Max Planck Society DE–80804 Munich (Germany) www.karger.com/psp Downloaded by: E-Mail dimitris_bolis @ psych.mpg.de lectics as a point of departure. The latter could be consid- ered as an evolving school of thought, met in various historical contexts (e.g., Greek, Chinese, Hegelian, and • Sociocultural Marxian dialectics [1, 2]) critical to both reductionism and dualism. It asserts that phenomena cannot be mean- ingfully understood by reducing them into single levels of description (cf. reductionism) or assuming a metaphysi- • Cognitive- cal independence between levels (cf. dualism), but should behavioral be rather studied in their wholeness, inner contradiction, and movement (Table 1). In this light, human mind and psychopathology cannot be understood in isolation from society, the body, and social interaction. To quote Hegel • Biological “to know, or, in other words, to comprehend an object is equivalent to being conscious of it as a concrete unity of opposed determinations” [3, 4]. We will, therefore, try to overcome traditional dichotomies, such as organism/en- vironment, by viewing them as both a result and a cause Fig. 1. Schematic depiction of dynamic interrelationships: between of reciprocal adjustments, or individual/society by con- multiple levels (e.g., biological, cognitive-behavioral, sociocultur- sidering the whole and the part as, albeit partially autono- al) and functions (e.g., including but not limited to the functional- mous, highly interdependent levels of organization. In ity of multiple neuromodulators or bacteria at the 1st level, body-/ this effort, we will also draw upon accounts of intersub- neurosystemic, and phenomenological aspects at the 2nd level, jectivity, which emphasize that single levels of analysis or and social structure, institutions, or cultural practices at the 3rd level) interacting in several temporal scales. Please note arrows cutting off the part from the whole may severely limit our may appear static on the image, but we interpret them as represen- understanding of a phenomenon. We will emphasize tations of developing interrelationships reflecting both quantita- viewing psychiatric conditions not as static conditions tive and qualitative changes (cf. dialectics). Certain additional core driven by a single cause, but rather as the outcome of an levels of description, i.e., the (micro-/macro-)physical levels, have interplay of multiple and diverse factors (Fig. 1) and to be been omitted from this illustration. more specific as a process of circular causality among dif- ferent levels of description (e.g., biological, cognitive-be- havioral, and sociocultural) as well as multiple functions within a level (e.g., action and perception within the cog- coarser mechanisms on a higher level might appear more nitive-behavioral level), unfolding over different tem- robust in terms of causality than relevant stochastic poral frames (e.g., evolutionary, cultural, social, individ- micro-processes. Thus, a genuine causal emergence on a ual-psychological, subindividual-biological developing macro-level is necessary for a complete description. In scales; based on Lev Vygotsky and colleagues’ views [5, 6] fact, this is a conclusion from physics where the circular on human development).1 causality between the microscopic and the macroscopic Indeed, contrary to a common assumption that a full is well established in terms of concepts such as the slaving description on a micro-spatiotemporal level is causally principle and the center manifold theorem. In brief, these complete, it has been suggested that a genuine causal theorems suggest the emergent macroscopic (order pa- emergence on a macro-level might also be possible [7]. rameters) that describe the whole enslave the microscop- Importantly, such an emergence is not to be solely attrib- ic components that constitute the whole. This induces a uted to a weakness of experimental means to fully grasp circular causality that lies at the heart of synergetics [cf. 8, the micro-phenomena but rather due to inherent charac- 148]. It also speaks to the circular causality to which en- teristics of systemic processes themselves. For example, activism and embodied (situated) cognition approaches appeal (Table 1). Following such a line of thought, this paper will argue that while considering neurobiological 1   Please note the specific definition and distinction between levels, func- and phenomenological processes is an important step to- tions, and temporal frames, as put forward here, are made for intelligibility ward the understanding of psychiatric conditions, it may purposes only and it should not be taken as implying dichotomies; processes and their interrelationships appear complex, continuous, and overlapping in remain incomplete as further levels of analysis, such as reality [145]. sociocultural processes and generally social structure, are 193.175.245.253 - 1/11/2018 5:02:03 PM 356 Psychopathology 2017;50:355–372 Bolis/Balsters/Wenderoth/Becchio/ DOI: 10.1159/000484353 Schilbach Max Planck Society Downloaded by: Table 1. Glossary of terms, as they either appear in the bibliography [5, 13, 68, 75, 149–154] or were introduced in this article Active inference An account of action according to which (biological) systems sample the environment in accordance with prior beliefs for minimizing free energy (see [13]) Bayesian account of Τhe “Bayesian account of intersubjectivity” is considered here as a Bayesian account of human activity that takes into intersubjectivity account both intra- and interpersonal processes (see this article) Biofeedback A training technique by which a person learns how to regulate certain body functions, such as heart rate, blood pressure, or brain wave patterns, that are normally considered to be involuntary (see [149]) Cultural-historical Theory aiming at accounting for the inseparable unity of mind, brain, and culture (see [5, 150]) psychology Dialectics The dialectical method states that phenomena can be understood only in their wholeness, inner contradiction, and movement (see [151]) Dialectical The “dialectical misattunement hypothesis” rethinks a psychiatric condition, such as autism spectrum conditions misattunement (ASC), not merely as a disorder of the individual brain but also as cumulative misattunement between persons, which can be thought of as disturbances in the dynamic and reciprocal unfolding of an interaction across multiple time scales, resulting in increasingly divergent prediction and (inter-)action styles (see this article) Dualism Theory stating that for some particular domain, there are two fundamental kinds or categories of things or principles (e.g., the physical and the mental) (see [152]) Emergence Emergent entities (properties or substances) “arise” out of more fundamental entities and yet are “novel” or “irreducible” with respect to them (see [153]) Free energy An information theory measure that bounds or limits (by being greater than) the surprise on sampling some data, given a generative model Put simply, with regards to an organism free energy minimization can be thought of as a process of maintaining current living form by being restricted in a limited number of possible states (see [13]) Heterogeneous dyads “Heterogeneous dyads” are considered here dyads which consist of persons with different conditions, such as a neurotypical person and a person with ASC (see this article) Homogeneous dyads “Homogeneous dyads” are considered here dyads which consist of either only neurotypical persons or only of persons with a certain condition, such as ASC (see this article) Interaction tuning “Interaction tuning” here refers to tuning of expectations of either or both interactors as well as facilitating a social interaction via tuning the communication medium (see this article) Intrapersonal (Processes) being unfolded within the person Interpersonal (Processes) being unfolded between persons Precision A statistical term defined as the inverse variance and can be thought of as the confidence a (biological) system places upon its beliefs (see [68, 75]) Prediction error The discrepancy between incoming information and (biological) system-generated predictions (see [68, 75]) Predictive coding/ Theory that states that (biological) systems are constantly generating and updating hypotheses about the causal processing structure of the environment and the self along different levels of abstraction for ultimately minimizing free energy (see [68, 75]) Sociofeedback “Sociofeedback” is considered here a (future) training technique by which a person, a dyad, or a group of people will learn how to (co-)regulate certain social interaction processes, such as interpersonal coupling and coordination; the concept also applies to automatic adjustment of the interaction medium based on social interaction monitoring (see this article) Synergetics An interdisciplinary field of research that studies the spontaneous, i.e., self-organized, formation of structures in systems far from thermal equilibrium (see [154]) 193.175.245.253 - 1/11/2018 5:02:03 PM The Dialectical Misattunement Psychopathology 2017;50:355–372 357 Hypothesis DOI: 10.1159/000484353 Max Planck Society Downloaded by: neglected. For instance, structures promoting social ex- to experience the environment in accordance with prior clusion or competitiveness, as opposed to communica- beliefs (i.e., active inference). Here, it should be noted that tion and collaboration, could distinctly shape individual Bayesian beliefs inherent in any Bayesian approach to behavior, mental reality, and biological mechanisms. cognition should largely be thought of as subpersonal. In Here, our approach heavily leans on work from the cul- other words, the experience subtended by predictive pro- tural historical activity theory (Table 1), which re-inter- cessing is not necessarily a conscious experience but more preted human development across a variety of conditions like a percept (or possibly a causative experience; i.e., qua- as a dynamic interplay between biological and sociocul- lia), embracing also other “automatic” processes such as tural forces ([6, 9, 10] on the work of Lev Vygotsky and homeostatic control. One of the many interesting aspects colleagues). Notably, the aforementioned variety of con- of this account is that perception, learning, and action are ditions was not limited to what one could think of “social not considered as isolated and passive processes, but they conditions” but rather included individuals who were constitute interconnected processes, which an organism both deaf and blind, to give an example. The organic con- actively deploys for making sense or (to put it in compu- dition can of course still affect the construction of the so- tational terms) “model” the world in order to maintain its cial self via atypical development if amelioration of social current living form [13]. exclusion is not taken into account. As Vygotsky, pointed out: The Dialectical Misattunement Hypothesis and a The confusion and failure to differentiate the organic from the Bayesian Account of Intersubjectivity cultural, the natural from the historical and the biological from the Taken together, we suggest that formally considering social […], inevitably leads to a fundamentally incorrect under- (both quantitative and qualitative) dynamically changing standing and interpretation of the data (observations) [excerpt interrelationships between and within levels of descrip- from Vygotsky’s work; translated in 6]. tion (Fig. 1) as well as temporal scales will be essential for a comprehensive understanding of complex psychiatric Psychiatry through a Computational Lens conditions, such as autism spectrum conditions (ASC). In our effort to adopt an integrative perspective, we In light of this, the purpose of this paper will be threefold: will use Bayesian accounts of cognition and behavior as Firstly, to consider the integration of diverse within-level powerful tools of analysis within the level of the individ- (i.e., neurocognitive) processes embedded in a common ual, but most importantly we will suggest ways of going framework, i.e., the predictive processing and active in- beyond the individual as the unit of analysis and eventu- ference account. Secondly, to outline the importance of ally overcoming limitations of a single-level approach taking into account interrelationships across levels (i.e., (see the last two chapters of this study). Computational the individual and the collective) via putting forward the psychiatry can be thought of as lying on the interface be- “dialectical misattunement” hypothesis. Thirdly, to ulti- tween computational neuroscience and clinical psychia- mately motivate the development of a “Bayesian account try. It deploys computational (e.g., Bayesian) modeling in of intersubjectivity” rather than of individual brains. Im- order to mechanistically describe psychiatric conditions portantly, we also highlight the practical implications of [11, 12]. A more specific hierarchical Bayesian approach our theoretical approach (i.e., ethical, research, clinical to perception and action, which we will focus on here, has and pedagogical). Taking ASC as a paradigm case, we will been described as the predictive coding (also mentioned give a description of the general framework of our ap- as predictive processing; a term which we will be using in proach. More concretely, we will first review the field of this article) and active inference account (Table 1). In autism research with emphasis on recent interest in pro- brief, according to such a perspective, the brain’s ultimate viding a Bayesian formulation of ASC. Based on this, we goal is the long-term minimization of free energy, which will argue in favor of adopting the Bayesian accounts of (as we will explain later under simplifying assumptions) brain function as a framework to integrate seemingly can be thought of as the “prediction error,” i.e., the dis- contradictory neurocognitive hypotheses. Then, we will crepancy between incoming information and generated discuss different accounts of intersubjectivity, which predictions, based on consolidated experience (Table 1). share a common ground by stating that individual level Importantly, this is thought to be accomplished through analyses do not suffice for a comprehensive understand- two main avenues, namely either via updating the beliefs ing of social perception and cognition. Bringing together one holds for aligning them with the environment (i.e., a dialectical perspective to human communication and predictive processing) or through action, which can help Bayesian (i.e., predictive processing and active inference) 193.175.245.253 - 1/11/2018 5:02:03 PM 358 Psychopathology 2017;50:355–372 Bolis/Balsters/Wenderoth/Becchio/ DOI: 10.1159/000484353 Schilbach Max Planck Society Downloaded by: accounts of individual mechanisms [14], we will intro- to any process evolving at the interface between the intra- duce the dialectical misattunement hypothesis which and interpersonal level (Table 1), including social exclu- emphasizes the interdependence of individual and collec- sion across different conditions. tive levels of description. More concretely, the dialectical misattunement hy- pothesis rethinks ASC not merely as a disorder of the Traditional Views on ASC individual brain but also as cumulative misattunement between persons. Misattunement across persons can be Although sparse references about resembling cases thought of as disturbances of the dynamic and recipro- may have existed before [17], it was not until the 1940s cal unfolding of an interaction across multiple time that Hans Asperger and Leo Kanner described the condi- scales, resulting in increasingly divergent prediction and tion of autism. Today, autism is considered as a neurode- (inter-)action styles. Consequently, with regard to neu- velopmental disorder spanning a spectrum characterized roscientific research, we propose moving from focusing by impairments in social interaction and communication only on comparing groups of individuals to considering as well as restricted, repeated behaviors and interests. It is types of interaction between persons (e.g., homogeneous also not uncommon for ASC individuals to show en- dyads consisted of either only neurotypical persons or hanced abilities for specific cognitive aspects including only of persons with a certain condition, as well as het- perception [18], attention [19], and memory [20]. While erogeneous dyads; including both tuned and nontuned some approaches have focused on the impairments, other interactions2, Table 1). Here the hypothesis holds clear accounts encompass both impaired and enhanced skills predictions: Interactions within homogeneous dyads [21, 22], especially when it comes to the so-called “high- are expected to appear smoother compared to heteroge- functioning” end of the spectrum. In the past half century, neous dyads. Additionally, tuned interactions of either a number of different cognitive hypotheses have been pur- homogeneous or heterogeneous dyads should appear as sued in order to understand core aspects of ASC. Although most effective. If these hypotheses are valid, the defini- several important ideas have helped to shed light on spe- tion of a psychiatric condition as ASC can be thought of cific facets, there is still no consensus about a single theo- as relative to the “other” and generally the social context. ry that could offer a universal and yet specific explanation Such an approach, will eventually allow us to escape an of the condition. We will primarily focus on the “5 big overly neurocentric research scope in psychiatry. Along ideas” about autism, as suggested by Uta Frith [23]: similar lines, we suggest that clinical and pedagogical Firstly, Baron-Cohen et al. [24] proposed that ASC in- practices should move beyond the individual to moni- dividuals lack a specific meta-representational capacity, toring, evaluating, and facilitating processes at the inter- namely a “theory of mind,” which prevents them from in- personal level. Also, reviewing ASC as a misattunement ferring upon other people’s mental states. As a conse- between people, and not as disorder of the brain per se, quence of this, ASC individuals – so it is assumed – can- may help to alleviate social stigma and reduce social ex- not know about other people’s beliefs, emotions, desires, clusion. perceptions, and intentions. In light of findings that ASC We will end by outlining a Bayesian account of inter- individuals can make a conscious effort to think about subjectivity, referred to as the “observing-the-interactors” others’ mental states, it has been suggested that implicit, scheme, which will allow us to computationally describe namely spontaneous, mechanisms of mentalizing might the interplay of individual and collective levels of activity be the ones that are primarily linked to relevant difficul- during social interactions. Subsequent papers will delin- ties in ASC, rather than explicit ones, which might be eate a practical approach for testing the misattunement easier compensated for through learning [25, 26]. hypothesis of social interaction based upon hierarchical The second big idea focuses on a special category of models of interpersonal interactions [15] and 2-person neurons, the so-called “mirror neurons” [27, 28], which psychophysiology [16]. In what follows, we focus on au- are active both when an action is performed and ob- tism, but the proposed approach more generally applies served. The broken mirror neuron (BMN) hypothesis pro- poses the explanation of impaired social skills in ASC on the basis of a dysfunctioning mirror neuron system 2 The term “tuned” here refers to multiple aspects: tuning expectations of   (MNS) [29, 30]. A number of studies offered supportive either or both the interactors, as well as facilitating the interaction via tuning the communication medium (e.g., social conventions, as well as the cultural evidence for the involvement of MNS [29, 31, 32]. How- or technological environment, in which the interaction is embedded). ever, both the validity of a broken MNS and a direct, caus- 193.175.245.253 - 1/11/2018 5:02:03 PM The Dialectical Misattunement Psychopathology 2017;50:355–372 359 Hypothesis DOI: 10.1159/000484353 Max Planck Society Downloaded by: al relationship between MNS and social skills in ASC have cific aspects of ASC; still, none of them is considered to been challenged by other reports [33, 34]. Differences in provide a global explanation. In fact, it has been argued MNS activation between neurotypical and ASC individu- that a single explanation at the cognitive, neural, or ge- als could be alternatively traced back to earlier modula- netic level might be intractable [51–53]. However, inter- tory effects of the mentalizing system as well [35, 36]. est in a potentially unifying account has recently re- Alternatively, the social motivation hypothesis focuses emerged while making reference to and drawing upon on motivational rather than ”purely cognitive” aspects the Bayesian brain hypothesis and particularly the pre- [37]. It proposes that people with ASC lack the inherent dictive processing and active inference scheme [16, 54– social drive, which would assist them in exploiting the 62]. In the following, we direct our attention to the dis- necessary learning opportunities for developing expertise cussion of this approach and its relevance for ASC. in social cognition. More precisely, the hypothesis is set- tled upon the fact that social orienting, social seeking and liking, as well as social maintaining appear to be affected Bayesian Approaches in ASC. On a biological level, the focus is placed on the human reward system, where either specific social im- The Bayesian Brain Hypothesis pairments or more general reward-related dysfunction The main premise of the Bayesian brain hypothesis could explain the behavioral findings. A suboptimal oxy- rests on the idea that the brain represents information tocin regulation has also been implicated in ASC, which accessed via the sensory organs in the form of proba- could, for example, reflect differences in relating social bility densities, as opposed to single numbers, which stimuli to rewarding values [38–41]. are continuously updated, as if following a specific set The fourth idea, namely the weak central coherence of mathematical formulas based on the Bayes theorem. hypothesis, considers ASC as a different, detail-oriented Crucially, this allows for optimal information integra- cognitive style [18, 42–44]. More precisely, it claims that tion both in time and space, multimodal cue integration, people with an ASC tend to process information locally as well as flexible information manipulation without the rather than globally. It predicts that people with ASC will need to commit to particular decisions at an early stage have difficulties in perceiving information in context. Ac- of processing [63]. To put it simply, through a Bayesian cording to this idea, ASC individuals perceive the world lens one can view the brain as an organ which calculates differently in a number of aspects, including visual, audi- and maintains probabilities about events in the environ- tory, and linguistic functions. Later, the enhanced percep- ment or about the self via a combination of already gained tual functioning hypothesis attributed this local bias to a experience and newly sensed information. Crucially, the superiority of detail processing per se and not due to in- more confidence (i.e., precision) is placed on the validity feriority of global information processing [45]. Mean- of experience (i.e., prior beliefs) the less the latter is up- while, the monotropism hypothesis proposed a general- dated in the face of new incoming information (i.e., evi- ization from the tendency to focus on a local level to a dence). need of focusing on a single source level of information To make it more intuitive, let us imagine a young [46]. woman, Penelope, living in Southern Greece, wakes up Finally, the executive dysfunction (ED) hypothesis fo- on a summer morning late for her work. The blinds are cuses on the difficulties that ASC individuals face when it shut down, and there is no time to check the weather out- comes to executive functions, i.e., problems primarily as- side the window. Will she take her umbrella on the way sociated with functions such as planning, flexibility, inhi- out? Based on her experience (i.e., prior beliefs: it rarely bition, and working memory [47–50]. For instance, dif- rains in Southern Greece in the summer), she decides not ficulties related to dealing with novel situations and im- to take her umbrella with her. However, in the evening provising, as well as perseverative stereotyped behavior in it happens to rain (evidence). The next day, Penelope, ASC, can be explained by ED. This hypothesis has been bringing together experience and the previous day’s facts, taken to suggest that the study of frontal cortex function thinks there might be a slightly higher probability of rain- should be particularly relevant for a neurofunctional un- ing (i.e., posterior belief), but this is still not high enough derstanding of ASC. to persuade her that carrying an umbrella might be a good To conclude this brief introduction of various ac- idea. After several days of raining, she eventually decides counts of ASC, it can be said that a number of different to put the umbrella in her bag. She has come to believe hypotheses have provided important insights into spe- that the probability of raining is high enough these days 193.175.245.253 - 1/11/2018 5:02:03 PM 360 Psychopathology 2017;50:355–372 Bolis/Balsters/Wenderoth/Becchio/ DOI: 10.1159/000484353 Schilbach Max Planck Society Downloaded by: despite her opposing experience of previous years. Per- biological mechanisms. Before explaining how a predic- haps not surprisingly from a Bayesian point of view, Pe- tive processing and active inference framework could, nelope still keeps the umbrella with her for a few days therefore, facilitate research into autism, we will first after the weather has been sunny and dry again. Before present the underlying basic ideas. concluding our example, it is worthwhile to introduce the concept of precision, which can be generally thought of Predictive Processing and Active Inference as the confidence about a certain belief. Let us imagine a The general idea of predictive processing and active second scenario, where Penelope wakes up on a summer inference is not new. For instance, one can find indica- morning in Japan, where she has been travelling for a few tions in Hermann von Helmholtz [70], who spoke about days. She has heard that the weather is generally dry in “unconscious inference” in the 19th century, drawing on summer in the city she stays. Yet, on the first day, it does ideas going back to ancient philosophers. Additionally, happen to rain. Interestingly, already from the next day, relevant traces can be found in ideas such as the reaffer- she decides to take an umbrella with her. Why did she ence and ideomotor principles [71–73]. To put it simply, change her mind so quickly in this case? Adopting a within a predictive processing and active inference frame- Bayesian perspective, one could argue that Penelope, al- work, the brain is essentially viewed as a “prediction though holding a high prior belief about not raining, machine” whose ultimate goal is the minimization of changes her mind quickly due to the relatively low confi- “prediction error” by deploying hierarchical generative dence (i.e., precision) she places on these prior beliefs of models. More precisely, higher levels of a hierarchy con- her, which have been the result of rumors and not her tinuously produce predictions, which are tested against own experience. the input information of the immediate lower levels. The discrepancy between predictions and incoming informa- The Hypo-Prior Hypothesis of Autism tion, i.e., the “prediction error,” is propagated to higher Coming back to our main example of ASC, Pellicano levels, reconfiguring the system to optimize its next pre- and Burr [54] adopted a Bayesian standpoint to argue that dictions. Notably, propagating only the error and not the nonsocial features of ASC might be well explained in ref- actual incoming information to higher levels is an effi- erence to attenuated Bayesian priors (i.e., priors of rela- cient and resource-oriented way of reducing the band- tively low precision, so-called hypo-priors). This hypoth- width of the processed information, which is also exploit- esis anticipates a relatively more “precise” perception in ed in data compression techniques, such as the common ASC, driven primarily by perceptual evidence as opposed JPEG format. In short, two processes take place at the to prior knowledge, as well as a sense of being over- same time in opposite directions; predictions are propa- whelmed by this information, a common complaint of gated backward from higher to lower levels, trying to ex- persons with ASC. Moreover, the hypo-prior hypothesis plain away prediction errors, and prediction errors are predicts the impedance of performance in ambiguous sit- propagated forward from lower to higher levels, updating uations when prior knowledge is crucial for optimally predictions (Fig.  2).3 The hierarchical structure of the solving a perceptual problem of inference. Finally, it was model is of immense importance because it enables the considered that a different learning style, namely one re- brain to optimize its own (empirical) priors on the fly. sembling overfitting in machine learning, and differences Additionally, it allows for effective representations of in- in adaptation can also be explained by this hypothesis creasing abstraction. From a neurobiological perspective, [cf. 64]. forward connections may arise in superficial pyramidal The hypo-prior hypothesis was then reformulated [56, cells, whereas the sources of backward connections are 57] within the predictive processing scheme, a more spe- assumed to reside in deep pyramidal cells [74, 75]. cific Bayesian account [65–68], while considering social At this point, it is important to place the predictive aspects of individual cognition and behavior [60, 61]. It is processing in the more general context of active inference worth noting that the importance of difficulties related to (a corollary of the free energy principle). Crucially, active predictions had been noted in the autism literature in the past as well [69]. However, the more recent shift towards focusing on predictive processing and particularly on the 3 Please note the new perspective, which is introduced with the predictive concept of precision described above can offer a poten- processing definitions of “backward” and “forward” connections, contrasted with the “feedback” and “feedforward” ones, since in the context of predic- tially unifying explanation of autistic symptoms and di- tive processing the backward connections are the ones providing feedback rectly relate computational findings with tractable neuro- via prediction error information on the forward stream of predictions [68]. 193.175.245.253 - 1/11/2018 5:02:03 PM The Dialectical Misattunement Psychopathology 2017;50:355–372 361 Hypothesis DOI: 10.1159/000484353 Max Planck Society Downloaded by:  Color version available online Fig. 2. A simplified representation of the predictive processing idea (taken from: Stefanics et. al. [146]): representation units (R; deep pyramidal cells) receive inputs (blue arrows) from error units (E; superfi- cial pyramidal cells) of the same (dotted line) and lower levels, while error units receive inputs (green arrows) from the same (dotted lines) and higher levels. Black arrows represent inhibitory intrinsic connections. inference takes predictive processing beyond the domain timizing these expectations for better matching with its of perceptual inference and provides an account of action. sensations (i.e., through perception and learning referred The brain can be seen as inferring upon the causal struc- to as predictive processing [76, 77]). More broadly, one ture of the world by updating “beliefs,” which are repre- could sketch a path which, starting from the existence of sented as probability densities. Most simply, the latter life (as a process leading to a restricted number of states), would take the form of Gaussian distributions, fully de- passes through entropy (referring to a tendency to resist fined by their mean (i.e., expectation) and variance (i.e., the 2nd law of thermodynamics), surprise (viewing en- inverse precision). Under this simplifying assumption tropy here as a mean value of surprise over time), free (i.e., the Laplace assumption), the generalization of pre- energy (as an upper bound of surprise), and eventually diction error minimization to “free energy” minimization leads to prediction error, which, as we pointed out, can be becomes mathematically more evident.4 The latter then considered as the free energy under certain simplifying takes the form of a difference between the predictions of assumptions. As provocatively put by Karl Friston [147] a model and the representations to be predicted [13]. In- “the motivation for minimizing free energy has hitherto deed, free energy had been originally formulated for con- used the following sort of argument: systems that do not fronting the difficult problem of exact inference, trans- minimize free energy cannot exist [...]” forming it into an easy problem of optimization. It could Crucially, in the setting of predictive processing and be possible that a similar trick is used by the brain in order active inference, the degree of prediction updating (i.e., to efficiently approximate the inference problem in a the learning rate) is controlled by the relative precision of quasi-optimal Bayesian way. Interestingly, the free energy successive levels. More precisely, it is proportional to a principle has been proposed as a potentially unifying relative precision-weighted prediction error. This makes brain theory, accounting for action, perception, and sense, since it would be generally desirable for an agent to learning. In short, an agent has two options for suppress- update their beliefs first when the prediction error is large ing free energy: first by selectively sampling the environ- and second when they are unsure (low precision or con- ment for fulfilling its own expectations (i.e., through fidence) about their prior beliefs compared to incoming acting referred to as active inference) and second by op- information of lower levels in the hierarchy [about the importance of precision, see 78]. Importantly, the idea of 4 an updating rule proportional to the precision-weighted   In this setting, free energy can be regarded as an approximation, namely an upper bound, to Bayesian model evidence, which is the probability of ob- prediction error is a potentially neurobiologically plau- serving the data given a specific model. sible account, where precision is assumed to be represent- 193.175.245.253 - 1/11/2018 5:02:03 PM 362 Psychopathology 2017;50:355–372 Bolis/Balsters/Wenderoth/Becchio/ DOI: 10.1159/000484353 Schilbach Max Planck Society Downloaded by: ed by the gain of superficial pyramidal cells calculating monly observed in ASC, could be viewed as efforts for precision errors [79–81]. Psychologically, increases and creating scenarios of reduced prediction error, because decreases in the precision of sensory prediction errors other pathways fail to do so. Finally, another core attri- have been associated with sensory attention and attenua- bute of ASC, i.e., withdrawal to one’s own self, might tion, respectively. In other words, attending to (or attenu- constitute an alternative strategy of generally keeping ating) a sensory stream is (under predictive processing) prediction errors low. This kind of behavior could also be mediated by affording more (or less) precision to that linked to an attenuation of motivational factors due to a stream [82]. persistent inefficiency to trigger reward through decreas- Before concluding this introduction to predictive pro- ing prediction errors [84, 85]. cessing and active inference, it is worth noting that this Intriguingly, certain predictions made by the aberrant scheme could be considered as a dialectical framework in precision hypothesis can be formally tested via deploying and of itself. Firstly, it defines action and perception as the predictive processing modeling. The latter approach al- interplay between two closely intertwined avenues for lows for the tracking of potentially critical processes of minimizing prediction error. New perceptual states can the hypothesized “predictive brain” and may, therefore, inform future actions, while informed adjustment and have the potential to become an invaluable tool for revis- sampling of the environment (i.e., action) decisively con- iting the condition of autism. To date, a number of differ- tributes to updating perception. Essentially, perception ent theoretical and computational predictive processing and action become here two dialectical facets of the same and active inference models have been put forward, cov- process, namely minimization of free energy. Addition- ering a variety of levels, functions, and temporal scales. In ally, prediction updating and activity can be viewed as the next section, we will suggest modeling examples of dialectical processes in time between prior experience potential relevance to the autism research at the individ- and incoming information, whose confrontation yields ual level. More specifically, we will view here predictive adjusted relations between environment and the self ei- processing and active inference as a common framework ther through updating current beliefs or the perceived for re-addressing traditional ideas about ASC. The “5 big environment itself. We again see here a circular causality ideas,” which rest on diverse cognitive functions, will mo- that is central to enactive (Bayesian) inference and speaks tivate and help to structure our suggestion. to related notions in enactivism and embodied cognition (see Integrating Individual and Collective Levels of Anal- ysis). After having provided a general introduction to the Individual Level: Predictive Processing and Active predictive processing and active inference framework, Inference as a Common Framework for Integrating their putative roles in understanding autism will be pre- Diverse Neurocognitive Hypotheses sented in the following. Theory of Mind – as described above – can be viewed as The Aberrant Precision Hypothesis of Autism an inference problem [86], where the brain tries to under- It has been suggested that considering the role of stand “invisible” mental states through observable human precision in cognitive and behavioral processes could be behavior. Koster-Hale and Saxe [87] review evidence that important for understanding differences between neuro- relates theory of mind to predictive processing formula- typical persons and ASC individuals: Indeed, there is pre- tions. To that end, they consider how relevant brain regions liminary neurobiological evidence with regard to the such as the superior temporal sulcus, temporoparietal junc- functionality of certain neuromodulators that is sugges- tion and medial prefrontal cortex might be involved in tive of aberrant precision in ASC [60]. Additionally, sev- mental state inference across different time scales. To be eral, psychological findings in ASC could be putatively more specific, the superior temporal sulcus has been impli- attributed to aberrant precision estimation [61, 83]. For cated in neural reactions to face and body action in the scale instance, hypersensitivity to sound and visual stimuli is of seconds, while the temporoparietal junction has been re- typically observed in ASC individuals [45]. Through a lated to assessing desires and beliefs of other people, which predictive processing and active inference lens, consider- can last from minutes to years, and the medial prefrontal ation of irrelevant information due to increased precision cortex has been thought to contribute to the evaluation of can possibly lead to perceptual overload or, in other temporally more stable traits of other people. words, perceptual hypersensitivity. Furthermore, stereo- The social motivation hypothesis of autism focuses on typies, repetitive behaviors, and self-stimulation, all com- how a lack of motivation for processing and learning 193.175.245.253 - 1/11/2018 5:02:03 PM The Dialectical Misattunement Psychopathology 2017;50:355–372 363 Hypothesis DOI: 10.1159/000484353 Max Planck Society Downloaded by: about social aspects could be relevant for understanding into the implications of a BMN account for understand- ASC or how difficulties in social cognition could decrease ing ASC. interest in social cues. Interestingly, Heyes [88] has ar- Visual processing and particularly the extraction of gued that social learning shares the same basic cognitive spatiotemporal regularities might also be related to spe- mechanisms with nonsocial learning. In line with this, cific theories about ASC, such as the weak central coher- Behrens et al. [89] indicated that standard reward-based ence hypothesis. Natural images tend to be correlated associative processes guide the acquisition of social infor- both in space and time. That is, natural scenes usually mation, too. More specifically, they showed activation of consist of finite regions of relatively uniform attributes the anterior cingulate cortex (ACC) gyrus and ACC sul- and tend to reflect region-specific uniform intensity val- cus for reward-based and social learning, respectively. At ues [97]. For example, a stable object, being viewed from the level of decision-making, it was found that the ventro- a constant perspective, appears to emit relatively similar medial prefrontal cortex encodes both probabilities about intensity values over time. These regular spatiotemporal social and nonsocial sources, appearing to integrate in- characteristics can be exploited by the visual system to formation from ACC sulcus and ACC gyrus in a subject- predict intensity values in advance based on neighbor- specific fashion. Consequently, the above-mentioned ing and historical information. Indeed, Rao and Ballard brain regions could potentially play an important role in [98] proposed that the brain predicts this kind of regu- the investigation of ASC-related differences in multi- larities via a predictive processing model embodied in modal cue integration and contextualization of precision neural loops of increasing receptive fields with ascend- in social and nonsocial cues [90, 91]. ing hierarchy (e.g., the lateral geniculate nucleus-prima- As previously discussed, the so-called “mirror neu- ry visual cortex-secondary visual cortex feedback loop ron system” has also been implicated in ASC via the [97]). Such a family of models could be exploited in the BMN hypothesis. According to the BMN hypothesis, future for an investigation of aspects related to a weak difficulties in ASC in understanding others’ actions and central coherence in ASC and more precisely the extrac- intentions may arise from a defective functioning of the tion of perceptual regularities. For instance, quantifying MNS. However, precisely how mirror neurons contrib- autism-specific styles in extracting such regularities ute to action/intention understanding is still unclear could yield further insights about facts as perceptual hy- [92]. Kilner et al. [93] suggested that the brain deploys a persensitivity and differences in perceiving certain kinds mirror neuron predictive processing model and mini- of illusion [99]. mizes prediction error at all levels. More specifically, The ED hypothesis focuses on executive cognition and they considered a hierarchy that consists of 4 levels of behavior. Kopp [100] has recently stressed the relevance decreasing abstraction descending the hierarchy; the (1) of executive function for predictive processing theories. intention, (2) goal, (3) kinematic level, and (4) muscular More precisely, drawing on the latter and self-terminat- level, respectively [94]. These levels of behavior are gen- ing operating units [101], Kopp proposed a theoretical erally assumed to be independent of each other [94]. hierarchical model for dealing with ED, especially focus- This assumption, however, appears not to be true as re- ing on brain regions as the medial, orbital, and lateral pre- cent evidence indicates that the kinematics of a per- frontal cortex. Indeed, there is evidence speaking for a formed movement already reflect the agent’s intention hierarchical organization of the rostrocaudal axis of the and makes it distinguishable [95]. This raises the in- prefrontal cortex based on the level of abstraction [102, triguing possibility that intentions may be decoded from 103]. We suggest such kind of models could prove to be movement kinematics [96]. A reasonable framework for fruitful in studying putative ED through the hierarchical integrating different sources of prediction is that a range inference entailed by predictive processing and active in- of possible intentions is first estimated from the spatial ference in ASC. and temporal context, e.g., in predictive areas outside Lawson et al. [60] have recently put forward several the mirror system [92]. This prior prediction can impact suggestions with regard to potentially aberrant predictive on action understanding, constraining the number of processing processes relevant for understanding ASC at a possible intentions. Early movement-discriminant ki- neurobiological level, too. For instance, plasma oxytocin, nematic features of the observed motor act can lead then which has been suggested to control the relative salience to the selection of the most probable intention. Studying of social and nonsocial stimuli [41], has been found to be such inference problems in light of predictive process- reduced in children with ASC [38]. These can be linked ing and active inference could provide further insights to an aberrant precision hypothesis under the assumption 193.175.245.253 - 1/11/2018 5:02:03 PM 364 Psychopathology 2017;50:355–372 Bolis/Balsters/Wenderoth/Becchio/ DOI: 10.1159/000484353 Schilbach Max Planck Society Downloaded by: that oxytocin is involved in contextualizing precision of counts of social cognition have been criticized for ne- social as compared to nonsocial stimuli [104]. glecting the interactive dimension of social situations and Taken together, we suggest that a multitude of aspects for adopting an individualistic view of (social) cognition in ASC can be integrated under the predictive processing (e.g., specifically on the example of autism [125], philo- and active inference perspective. By doing so, ASC can be sophical considerations [107], and neuroscientific re- revisited as a different prediction and (inter-)action style, search [120]). With regard to psychiatric conditions, it as opposed to a set of a priori impaired neurocognitive has also been suggested that transdiagnostically observed functions that reside in specific brain regions. This exact social impairments are more likely or may only manifest shift of perspective, however, begs the question of how under conditions of real-time social interaction, whereas does such a different style emerge? In the next section, we situations of social observation might be less problematic tackle this question by leaning on sociocultural historical [123]. Furthermore, several accounts have been critical theories, which emphasize the social construction of the toward core assumptions of contemporary cognitivist (a-)typical self, and Bayesian accounts of brain function, paradigms, which have been thought of as viewing the which provide a powerful toolbox for the investigation of brain, or more generally the organism, merely as a passive underlying mechanisms. “consumer” of external stimuli [126]. Despite each ac- count’s distinct commitments, these kinds of approaches are usually positioned under the umbrella of the “4Es” Integrating Individual and Collective Levels of [127, 128], which described cognition as enactive [129– Analysis: The Dialectical Misattunement Hypothesis 131], embodied [132–134], embedded [130, 132, 133], and extended [130, 135], but also affective [136, 137]. In We open this section by discussing different ap- line with these accounts, using scenarios of higher eco- proaches which – although following distinct lines of ar- logical validity, which do not neglect the critical role of gument – converge on the idea that focusing on individ- the body, the environment and interactions in cognition ual brains will not be enough to fully understand the hu- could offer a more suitable framework to study brain man mind and psychopathology. In particular, we will function and behavior [16, 120]. argue against considering only biological mechanisms, On top of providing a naturalistic scenario, interactive since, in our view, the latter reductionist approach covers situations also potentially allow for the consideration of only part of the dialectical interplay between individual turn taking [112] and emergent social phenomena at processes and the collective level. In fact, cultural histori- higher levels of description, which otherwise might re- cal activity theories have strongly emphasized the impor- main intangible [15]. In neuroscience, cognition has gen- tance of considering the interrelationship between indi- erally tended to imply a dynamic interaction between vidual and sociocultural processes in psychological and brain areas merely within a single skull. However, there is psychopathological research: For instance, Vygotsky al- no theoretical reason to a priori exclude other body parts, ready distinguished social interaction as a key factor in and generally other people, as well as mediating cultural the formation of consciousness and “higher” human psy- tools, as cultural historical activity theories would empha- chological processes, which he argued are developed size. In line with an enactivist or dynamical system per- through and due to social interactions [6]. Additionally, he spective, two or more communicating agents can be seen claimed that every function appears twice in a child’s de- as a coupled system, being driven by nonlinear interac- velopment, first at a social level (i.e., “intermind”) and tions [113, 114, 138]. However, investigating individual then at an individual level (i.e., “intramind”): “All the predictive processing mechanisms in order to understand higher functions originate as actual relationships between communicative processes between agents could also be individuals” [5]. In other words, he suggested that through particularly informative. Notably, a formal account of ad- communication, through the direct social interaction dressing communication as reciprocal exchange of pre- with others, a child internalizes active cultural values in dictions about the other’s behavior has recently been put society [as cited in 6], realizing that the (a-)typical self is forward [139, 140]. This account, which rests on predic- dialectically and socially constructed. tive processing, considers both perceptual updating and Interestingly, recent developments in accounts of so- action expression within a closed loop between two cial cognition and intersubjectivity have also focused on agents. Here, simulations were used to illustrate how two the enabling or even constitutive role of social interaction agents, which model each other, could in theory converge [15, 16, 95, 105–124]. More specifically, mainstream ac- into a system of generalized synchrony (i.e., synchroniza- 193.175.245.253 - 1/11/2018 5:02:03 PM The Dialectical Misattunement Psychopathology 2017;50:355–372 365 Hypothesis DOI: 10.1159/000484353 Max Planck Society Downloaded by:  Color version available online typical self. Consequently, the interactive nature of so- cial situations can help to enhance or decrease differ- Intersubjective communicative gap ences in prediction and (inter-)action style in a feedback loop fashion (cf. the circular causality introduced above). That is, small initial differences at the individual level are thought of cumulatively enhancing (or weakening) Individual prediction styles interpersonal coupling during social interaction and vice versa. Schematically, an initial communicative gap could yield incompatible prediction and (inter-)action styles and vice versa (Fig. 3). Notably, such communicative misattunement could Fig. 3. Dialectical misattunement: increasing communicative gap (collective level) yields increasingly different prediction and be expected to unfold across multiple temporal scales; for (inter-)action styles (individual level) and vice versa. example, this could take place during the course of a dia- logue (scale of minutes), during a human relationship (scale of months or years), or along development (scale of a lifetime). Additionally, with regard to groups of people tion of chaos), thereby effectively embodying a single (e.g., the so-called psychopathological groups or, gener- shared model. In contrast to this ‘solipsistic’ understand- ally, any other social group), this kind of misattunement ing of communication, we argue that by adopting a dia- could even take on a cultural form, spanning a scale of lectical perspective we will look for such synchronization several generations. For instance, culturally cultivated be- dynamics across different levels of description and do not liefs in a given society about a specific group of people assume that my understanding of another is realized en- (e.g., stereotypes) might modulate the communication ef- tirely in my own head. ficacy between in- and out-group persons. To be more specific, we suggest that a “dialectical More broadly, we believe that for gaining a complete misattunement” constitutes one of the defining factors understanding of conditions such as ASC, a shift in focus of ASC and other psychiatric conditions. Communica- from the individual brain to the interaction between peo- tion misalignments and weak interpersonal coupling in ple is essential. Intriguingly, as we will argue in the next social interactions might be the result of increasingly di- and final section, such an approach could yield formal vergent predictive and (inter-)action styles across indi- insights into both individual and collective mechanisms viduals (cf. Predictive Processing and Active Inference). [15], as well as intra- and intercondition communication From an ontogenetic perspective, such a misattunement characteristics. Additionally, in psychiatry, it could facil- could result in impoverished opportunities for acquir- itate research at both diagnostic and treatment levels. In ing socioculturally mediated knowledge and skills. In short, we view the future of relevant theoretical research other words, we view two potentially cardinal processes and clinical practice not only as an investigation of “dis- that are tightly intertwined in a dialectical relationship: ordered” brain mechanisms but also of a “misattunement” at the collective level weak coupling, crucially modulated between persons. In line with the dialectical misat- by sociocultural factors, might lead to greater interindi- tunement hypothesis, which highlights intersubjectivity vidual incompatibilities in prediction and (inter-)action as an indispensable factor of human development, we also styles, while at the individual level, diverging prediction suggest the enrichment of approaches which exclusively and (inter-)action styles might lead to weak communica- aim at “tuning” the ASC person. To this end, we suggest tive coupling with others in social interaction.5 In short, considering tuning also the “other” (i.e., the neurotypical “dialectical misattunement” refers to an imbalance be- person with whom the ASC person interacts), as well as tween individual and collective levels rather than exclu- the social interaction medium (i.e., sociocultural frame- sively considering single levels. This view particularly work, such as social expectations and stereotypes, as well highlights the critical role of social interaction into hu- as the technological medium, such as educational social man development and the social construction of the (a-) robotics) [16]. More precisely, in a clinical setting, one could, there- 5 fore, pay attention not only to the potentially “maladap- Please note misattunement encompasses both aspects of dissimilari- ty (e.g., social misalignment) and noncomplementarity (e.g., dysregulated tive” processes within the diseased individual but also to coupling). the coupling dynamics of the dyad (for instance during 193.175.245.253 - 1/11/2018 5:02:03 PM 366 Psychopathology 2017;50:355–372 Bolis/Balsters/Wenderoth/Becchio/ DOI: 10.1159/000484353 Schilbach Max Planck Society Downloaded by:  Color version available online Individual tuning Individual with autism Interaction tuning Tuned dyad Individual without autism Individual tuning Fig. 4. Schematic presentation of a misattunement amelioration: by intervening both at the individual (e.g., cog- nitive and behavioral training of both interactors) and the collective level (e.g., adjustments of cultural/techno- logical tools, sociofeedback). Blue, individual trajectory of a person with autism spectrum conditions (ASC); orange, individual trajectory of a person without ASC (trajectories here represent multiple temporal scales, from minutes in the course of a conversation to years across development). psychotherapy or group sessions) and critically the inter- cially, while biofeedback techniques have been fruitfully action between the individual and the collective. Addi- used for monitoring and constructively exploiting indi- tionally, our approach also motivates an alternative peda- vidual activity (e.g., physiological factors), our approach gogical program. The latter would primarily aim at tun- would further point toward an extended notion of feed- ing not merely individual behavior but crucially the back, here referred to as “sociofeedback” (Table 1), includ- interaction between people. Here, the pedagogical proce- ing relational parameters (e.g., interpersonal coupling), dure would move beyond the traditional classroom, fo- too. Furthermore, the proposed shift in attention could cusing on cognitive and behavioral aspects of not only not only be beneficial in clinical and pedagogical practice the person with a specific condition (e.g., ASC) but also but also more broadly with regard to societal practice. their interactors (e.g., parents, educators, or peers) and, For instance, by diffusing ideas in society about view- most importantly, communication and mediating factors ing psychiatric conditions as disorders of social interac- (Fig. 4). tion [123] rather than disorders of individuals, psychiat- This could be achieved by developing adjustable ric stigma could be attenuated. As Vygotsky used to frameworks both to the individual and the interaction it- highlight, simply speaking, aspects of specific difficulties self. A promising solution could be found in the form of related to psychiatric conditions can be thought of as fall- “smart” technology, which could track and guide tradi- ing into two main categories: first aspects which are di- tional educational practice, taking into account real-time rectly related to a biological level and second aspects activity but also historically relevant aspects [141]. Cru- which are related to relevant beliefs and practices in soci- 193.175.245.253 - 1/11/2018 5:02:03 PM The Dialectical Misattunement Psychopathology 2017;50:355–372 367 Hypothesis DOI: 10.1159/000484353 Max Planck Society Downloaded by: ety. Although social processes play a decisive role in shap- revisited as a different prediction and (inter-)action style, ing a person’s mental reality, emphasis is usually only giv- as opposed to a set of a priori impaired neurocognitive en to biology. Notably, being a social product to a large functions that reside in specific brain regions. Then, we extent, such difficulties could be historically (along both argued that such an approach is not sufficient on its own social-historical and individual-developmental trajecto- but needs to be directed towards the relevant real-life ries) alleviated. Furthermore, our approach emphasizes phenomena that take place during social interaction. the dialectical relation of the collective and the individual Consequently, we propose an approach for integrating a (e.g., interrelations between culture and individual computational and a dialectical perspective to psychiatric persons, as in interactions between “patient” and “exam- conditions for scientifically studying both intra- and iner,” or “patient” and “non-patient”). The broadened interpersonal processes by introducing the “dialectical scope of effective treatment could encompass both per- misattunement” hypothesis. Misattunement across per- sonal and interpersonal parameters. In this light, the rela- sons is thought of as disturbances in the dynamic and re- tivity of psychiatric diagnosis, which is usually the out- ciprocal unfolding of an interaction across multiple time come of a communicative procedure between a potential scales, resulting in increasingly divergent prediction and patient and a culturally tuned examiner (e.g., psychiatrist (inter-)action styles (ways of generating and expressing or psychologist), also becomes more evident [10, 16, 115, expectations about the [social] world and the self). This 123, 142]. In technical terms, our approach could be re- thesis does not consider psychiatric conditions, such as framed as studying potential dynamic and recurrent feed- ASC, merely as disordered function within individual back loops across and within different levels of descrip- brains but rather as an interactive mismatch between per- tion, as well as temporal scales, driving both quantitative sons. and qualitative changes (cf. dialectics). We believe that In a forthcoming paper, we will use the conceptual computational modeling, such as Bayesian accounts, as arguments introduced above to illustrate the dialectical well as dynamical system approaches can prove to be misattunement hypothesis formally. Specifically, we will fruitful tools for scientifically testing the potentials of analyze two-person simulations and experiments [16] such a perspective. In fact, in our closing section, we will with dual hierarchical Gaussian filters [143] as a formal motivate a Bayesian account of intersubjectivity, which (computational) model of dyadic exchange [15]. This pro- will aim at formally accommodating both individual and vides a quantitative and principled description of the dia- collective mechanisms. lectical misattunement hypothesis, and how it could be verified empirically using relatively simple paradigms and analyses. In concrete terms, we suggest that established Summary and Outlook: From a Synthesis of techniques of multilevel computational modeling [143, Dialectical and Computational Approaches to a 144] can be used to investigate the interrelation of indi- Bayesian Account of Intersubjectivity vidual brain mechanisms and interpersonal processes. In- trasubjective parameters (e.g., on the dynamics of belief In this article, taking dialectics as a point of departure updating) will be used for modeling individual brain pro- and drawing upon insights from multiple areas of re- cesses of two (or more) brains, while intersubjective pa- search, we have argued that considering inherent inter- rameters will be introduced on a second meta-Bayesian relations as well as integrating findings from diverse lev- level for capturing dyadic (or group collective) processes, els of description, within-level processes and multiple such as interpersonal coupling [15]. The latter scheme will temporal scales will be essential in future autism research. thus move beyond current neuromodeling approaches by Such a holistic development, we claim, will help to unveil also considering emergent phenomena on higher levels of the intrinsic units of analysis for reconstructing the criti- description, such as questions about the autonomy of a cal dimensions of a multilevel and multidimensional dyad or a group of people and the individuality of the condition such as ASC: thus, it is here thought of as an mind. To give a more specific example, in the context of “autism space” rather than a spectrum. In particular, we collective decision-making or joint action, a nonlinear discussed how a framework such as predictive processing model might optimally explain observed behavior, thus, and active inference could be used to bring traditional providing evidence that the dyad or the group is different hypotheses at the level of the individual (e.g., neurobiol- than the sum of individuals. Inversely, this framework ogy, cognition, and behavior) together and re-address could address questions about how mechanisms of soci- them under a common umbrella. By doing so, ASC was etal structure and, in general, collective processes, in turn, 193.175.245.253 - 1/11/2018 5:02:03 PM 368 Psychopathology 2017;50:355–372 Bolis/Balsters/Wenderoth/Becchio/ DOI: 10.1159/000484353 Schilbach Max Planck Society Downloaded by: shape individual reality. For instance, one could differen- which will embrace the individual with autism as well as tially study the potentially distinct impact which a com- the socioculturally mediated interactions with other peo- petitive versus a collaborative structure might exert upon ple. The ultimate goal of such an approach will be to go an individual. Notably, this kind of modeling architecture beyond current diagnostic and treatment practice by pro- will not be merely able to model multiple levels of descrip- moting a reciprocal alignment of individual and societal tion but interlevel processes as well (e.g., internalization practices as opposed to a single-sided adjustment of indi- and externalization mechanisms). vidual behavior and brain function into the “normal”. Moving the focus from the observation of individual observers toward a multilevel observation of dyads and groups of interactors could help to explore whether and Acknowledgments how interpersonal coordination might actually serve as a prior and modulate the need for inferences about hidden Dimitris Bolis would like to express his gratitude to colleagues in the Translational Neuromodeling Unit (ETH Zurich), Autism causes of social behavior. Such an intersubjectively Bayes- Research Group (Trinity College Dublin), and the Independent ian approach, we claim, will provide a formal character- Max Planck Research Group for Social Neuroscience (Max Planck ization of subject-specific as well as dyad and group level Institute of Psychiatry). In particular, Dimitris is grateful to Klaas dynamics. It will, thereby, significantly advance our un- Enno Stephan and Karl Friston, as well as two anonymous review- derstanding of ASC and other psychiatric conditions ers for helpful feedback, but would also like to warmly thank Ky- veli Kompatsiari, Andreas Milias, Falk Lieder, Codruta Sudrijan, thought of as disorders of social interaction. As we pro- Angeliki Dotsi, and Leontios Hadjileontiadis, for inspiring discus- vocatively state in the title of this article, we suggest we sions on a multitude of topics from Vygotsky and Bayes to dialec- need to go beyond autism – not by neglecting the exis- tics, enactivism, and autism. tence of the condition but by adopting a holistic approach References 1 Wong W: Understanding dialectical thinking 12 Huys QJM, Maia TV, Frank MJ: Computa- 19 Plaisted Grant K, Davis G: Perception and ap- from a cultural-historical perspective. Philos tional psychiatry as a bridge from neurosci- perception in autism: rejecting the inverse as- Psychol 2006;19:239–260. ence to clinical applications. Nat Neurosci sumption. Philos Trans R Soc B Biol Sci 2009; 2 Dafermos M: Rethinking the crisis in social 2016;19:404–413. 364:1393–1398. psychology: a dialectical perspective. Soc 13 Friston K: The free-energy principle: a unified 20 Treffert DA: The savant syndrome: an ex- Personal Psychol Compass 2015;9:394–405. brain theory? Nat Rev Neurosci 2010;11:127– traordinary condition. A synopsis: past, pres- 3 Bidell T: Vygotsky, Piaget and the dialectic of 138. ent, future. Philos Trans R Soc London B Biol development. Hum Dev 1988;31:329–348. 14 Bolis D, Becchio C, Schilbach L: Revisiting Sci 2009;364:1351-1357. 4 Lawler J: Dialectical philosophy and develop- psychological definitions at the interface of 21 Frith U, Happé F: Autism: beyond “theory of mental psychology: Hegel and Piaget on con- sociocultural historical theories and predic- mind” (review). Cognition 1994;50:115–132. tradiction. Hum Dev 1975;18:1–17. tive coding; in Worlding the Brain: Patterns, 22 Baron-Cohen S: Is Asperger syndrome/high- 5 Vygotsky LS: Mind in Society: The Develop- Rhythms, Narratives in Neuroscience and the functioning autism necessarily a disability? ment of Higher Psychological Processes. Humanities. Amsterdam, 2016, p 10. https:// Dev Psychopathol 2000;12:489–500. Cambridge, Harvard University Press, 1930– worldingthebrain2016.files.wordpress.com/ 23 Frith U: Autism: A Very Short Introduction. 1935/1980. 2015/12/booklet-wtb.pdf. Oxford, Oxford University Press, 2008. 6 Vagenas E: The Soviet Psychology for the De- 15 Bolis D, Schilbach L: Beyond one Bayesian 24 Baron-Cohen S, Leslie AM, Frith U: Does the velopment of the Child. Athens, Synchroni brain: modeling intra- and inter-personal autistic child have a “theory of mind?” Cogni- Epochi, 2003. processes during social interaction. Com- tion 1985;21:37–46. 7 Hoel EP, Albantakis L, Tononi G: Quantify- mentary on “Mentalizing homeostasis: the 25 Senju A, Southgate V, White S, Frith U: Mind- ing causal emergence shows that macro can social origins of interoceptive inference” by blind eyes: an absence of spontaneous theory beat micro. Proc Natl Acad Sci USA 2013;110: Fotopoulou & Tsakiris. Neuropsychoanalysis of mind in Asperger syndrome. Science 2009; 19790–19795. 2017;19:35–38. 325:883–885. 8 Haken H: Synergetics: Introduction and Ad- 16 Bolis D, Schilbach L: Observing and partici- 26 Schilbach L, Eickhoff SB, Cieslik EC, Kuzma- vanced Topics. Berlin, Springer, 2004. pating in social interactions: action percep- novic B, Vogeley K: Shall we do this together? 9 Vygotsky LS: Thought and language. Athens, tion and action control across the autistic Social gaze influences action control in a Gnosi (translated in Greek by Maria and An- spectrum. Dev Cogn Neurosci DOI: 10.1016/j. comparison group, but not in individuals gelina Rodi), 1934/1964. dcn.2017.01.009. with high-functioning autism. Autism 2012; 10 Dafermos M: The Cultural Historical Theory 17 Wing L: The history of ideas on autism. 16:151–162. of Vygotsky. Athens, Atrapos, 2002. Autism 1997;1:13–23. 27 di Pellegrino G, Fadiga L, Fogassi L, Gallese V, 11 Stephan KE, Mathys C: Computational ap- 18 Dakin S, Frith U: Vagaries of visual percep- Rizzolatti G: Understanding motor events: proaches to psychiatry. Curr Opin Neurobiol tion in autism. Neuron 2005;48:497–507. a neurophysiological study. Exp Brain Res 2014;25:85–92. 1992;91:176–180. 193.175.245.253 - 1/11/2018 5:02:03 PM The Dialectical Misattunement Psychopathology 2017;50:355–372 369 Hypothesis DOI: 10.1159/000484353 Max Planck Society Downloaded by: 28 Rizzolatti G, Craighero L: The mirror-neuron 46 Murray D, Lesser M, Lawson W: Attention, 64 Harris H, Israeli D, Minshew N, Bonneh Y, system. Annu Rev Neurosci 2004;27:169–192. monotropism and the diagnostic criteria for Heeger DJ, Behrmann M, et al: Perceptual 29 Oberman LM, Hubbard EM, McCleery JP, autism. Autism 2005;9:139–156. learning in autism: over-specificity and pos- Altschuler EL, Ramachandran VS, Pineda JA: 47 Steel MJG, Gorman R, Flexman JE: Neuro- sible remedies. Nat Neurosci 2015; 18: 1574– EEG evidence for mirror neuron dysfunction psychiatric testing in an autistic mathematical 1576. in autism spectrum disorders. Cogn Brain idiot-savant: evidence for nonverbal abstract 65 Mumford D: On the computational architec- Res 2005;24:190–198. capacity. J Am Acad Child Psychiatry 1984; ture of the neocortex. The role of cortico-cor- 30 Ramachandran VS, Oberman LM: Broken 23:704–707. tical loops. Biol Cybern 1992;66:241–251. mirrors: a theory of autism. Sci Am 2006;295: 48 Rumsey JM: Conceptual problem-solving in 66 Friston K: A theory of cortical responses. 62–69. highly verbal, nonretarded autistic men. J Philos Trans R Soc Lond B Biol Sci 2005;360: 31 Dapretto M, Davies MS, Pfeifer JH, Scott AA, Autism Dev Disord 1985;15:23–36. 815–836. Sigman M, Bookheimer SY, et al: Under- 49 Ozonoff S, Pennington BF, Rogers SJ: Execu- 67 Friston K: Hierarchical models in the brain. standing emotions in others: mirror neuron tive function deficits in high-functioning au- PLoS Comput Biol 2008;4:e1000211. dysfunction in children with autism spectrum tistic individuals: relationship to theory of 68 Clark A: Whatever next? Predictive brains, disorders. Nat Neurosci 2006;9:28–30. mind. J Child Psychol Psychiatry 1991; 32: situated agents, and the future of cognitive 32 Perkins T, Stokes M, McGillivray J, Bittar R: 1081–1105. science. Behav Brain Sci 2013;36:181–204. Mirror neuron dysfunction in autism spec- 50 Hill EL: Executive dysfunction in autism. 69 Gomot M, Wicker B: A challenging, unpre- trum disorders. J Clin Neurosci 2010; 17: Trends Cogn Sci 2004;8:26–32. dictable world for people with autism spec- 1239–1243. 51 Happé F: Cognition in autism: one deficit or trum disorder. Int J Psychophysiol 2012; 83: 33 Southgate V, Hamilton AF: Unbroken mir- many; in Bock G, Goode J (eds): Autism: Neu- 240–247. rors: challenging a theory of autism. Trends ral Basis and Treatment Possibilities: Novartis 70 von Helmholtz HLF: Handbuch der physiolo- Cogn Sci 2008;12:225–229. Foundation Symposium 251. Chichester, Wi- gischen Optik. Leipzig, Voss, 1867. 34 Fan Y-T, Decety J, Yang C-Y, Liu J-L, Cheng ley, 2003, pp 198–212. 71 James W: The Principles of Psychology. New Y: Unbroken mirror neurons in autism spec- 52 Happé F, Ronald A, Plomin R: Time to give York, Holt, 1890. trum disorders. J Child Psychol Psychiatry up on a single explanation for autism. Nat 72 Stock A, Stock C: A short history of ideo- 2010;51:981–988. Neurosci 2006;9:1218–1220. motor action. Psychol Res 2004;68:176–188. 35 Wang Y, Hamilton AF: Social top-down re- 53 Gallagher S, Varga S: Conceptual issues in au- 73 Herbort O, Butz MV: Too good to be true? sponse modulation (STORM): a model of the tism spectrum disorders. Curr Opin Psychia- Ideomotor theory from a computational per- control of mimicry in social interaction. Front try 2015;28:1. spective. Front Psychol 2012;3:494. Hum Neurosci 2012;6:153. 54 Pellicano E, Burr D: When the world becomes 74 Felleman DJ, Van Essen DC: Distributed hi- 36 Dumas G, Soussignan R, Hugueville L, Mar- “too real”: a Bayesian explanation of autistic erarchical processing in the primate cerebral tinerie J, Nadel J: Revisiting mu suppression perception. Trends Cogn Sci 2012;16:504–510. cortex (review). Cereb Cortex 1991;1:1–47. in autism spectrum disorder. Brain Res 2014; 55 Brock J: Alternative Bayesian accounts of au- 75 Friston K, Kiebel S: Predictive coding under 1585:108–119. tistic perception: comment on Pellicano and the free-energy principle. Philos Trans R Soc 37 Chevallier C, Kohls G, Troiani V, Brodkin ES, Burr. Trends Cogn Sci 2012;16:573–574. B Biol Sci 2009;364:1211–1221. Schultz RT: The social motivation theory of 56 Friston KJ, Lawson R, Frith CD: On hyperpri- 76 Friston KJ, Stephan KE: Free-energy and the autism. Trends Cogn Sci 2012;16:231–239. ors and hypopriors: comment on Pellicano brain. Synthese 2007;159:417–458. 38 Modahl C, Green L, Fein D, Morris M, Wa- and Burr. Trends Cogn Sci 2013;17:1. 77 Friston KJ, Daunizeau J, Kilner J, Kiebel SJ: terhouse L, Feinstein C, et al: Plasma oxytocin 57 van Boxtel JJA, Lu H: A predictive coding per- Action and behavior: a free-energy formula- levels in autistic children. Biol Psychiatry spective on autism spectrum disorders. Front tion. Biol Cybern 2010;102:227–260. 1998;43:270–277. Psychol 2013;4:19. 78 Clark A: The many faces of precision (Replies 39 Bartz JA, Zaki J, Bolger N, Ochsner KN: Social 58 Van de Cruys S, de-Wit L, Evers K, Boets B, to commentaries on “Whatever next? Neural effects of oxytocin in humans: context and per- Wagemans J: Weak priors versus overfitting prediction, situated agents, and the future of son matter. Trends Cogn Sci 2011;15:301–309. of predictions in autism: reply to Pellicano cognitive science”). Front Psychol 2013;4:270. 40 Modi ME, Young LJ: The oxytocin system in and Burr (TICS, 2012). Iperception 2013; 4: 79 Feldman H, Friston KJ: Attention, uncertain- drug discovery for autism: animal models and 95–97. ty, and free-energy. Front Hum Neurosci novel therapeutic strategies. Horm Behav 59 Teufel C, Subramaniam N, Fletcher PC: The 2010;4:215. 2012;61:340–350. role of priors in Bayesian models of percep- 80 Shipp S, Adams RA, Friston KJ: Reflections 41 Gordon I, Vander Wyk BC, Bennett RH, tion. Front Comput Neurosci 2013;7:25. on a granular architecture: predictive coding Cordeaux C, Lucas MV, Eilbott JA, et al: Oxy- 60 Lawson RP, Rees G, Friston KJ: An aberrant in the motor cortex. Trends Neurosci 2013; tocin enhances brain function in children precision account of autism. Front Hum Neu- 36:706–716. with autism. Proc Natl Acad Sci USA 2013; rosci 2014;8:302. 81 Moran RJ, Campo P, Symmonds M, Stephan 110:20953–20958. 61 Van de Cruys S, Evers K, Van der Hallen R, KE, Dolan RJ, Friston KJ: Free energy, precision 42 Frith U: Autism: Explaining the Enigma. Van Eylen L, Boets B, De-Wit L, et al: Precise and learning: the role of cholinergic neuromod- Oxford, Blackwell, 2003. minds in uncertain worlds: predictive coding ulation. J Neurosci 2013;33:8227–8236. 43 Happé F: Autism: cognitive deficit or cogni- in autism. Psychol Rev 2014;121:649–675. 82 Friston K: The free-energy principle: a rough tive style? Trends Cogn Sci 1999;3:216–222. 62 Sevgi M, Diaconescu AO, Tittgemeyer M, guide to the brain? Trends Cogn Sci 2009;13: 44 Happé F, Frith U: The weak coherence ac- Schilbach L: Social Bayes: using Bayesian 293–301. count: detail-focused cognitive style in autism modeling to study autistic trait-related differ- 83 Hohwy J: The Predictive Mind. Oxford, Ox- spectrum disorders. J Autism Dev Disord ences in social cognition. Biol Psychiatry ford University Press, 2014. 2006;36:5–25. 2016;80:112–119. 84 Van de Cruys S, Wagemans J: Putting reward 45 Mottron L, Dawson M, Soulières I, Hubert B, 63 Knill DC, Pouget A: The Bayesian brain: the in art: a tentative prediction error account of Burack J: Enhanced perceptual functioning in role of uncertainty in neural coding and com- visual art. Iperception 2011;2:1035–1062. autism: an update, and eight principles of au- putation. Trends Neurosci 2004;27:712–719. 85 Joffily M, Coricelli G, FitzGerald T, Galea JM, tistic perception. J Autism Dev Disord 2006; Adams R: Emotional valence and the free- 36:27–43. energy principle. PLoS Comput Biol 2013; 9:e1003094. 193.175.245.253 - 1/11/2018 5:02:03 PM 370 Psychopathology 2017;50:355–372 Bolis/Balsters/Wenderoth/Becchio/ DOI: 10.1159/000484353 Schilbach Max Planck Society Downloaded by:  86 Baker CL, Saxe RR, Tenenbaum JB: Bayesian 105 Gallagher S: The practice of mind: theory, 124 Koike T, Tanabe HC, Sadato N: Hyperscan- theory of mind: modeling joint belief-desire simulation or primary interaction? J Con- ning neuroimaging technique to reveal the attribution. Proc 32nd Annu Conf Cogn Sci scious Stud 2001;8:83–108. “two-in-one” system in social interactions. Soc, Boston, 2011. 106 Schilbach L, Wohlschlaeger AM, Kraemer Neurosci Res 2015;90:25–32. 87 Koster-Hale J, Saxe R: Theory of mind: a NC, Newen A, Shah NJ, Fink GR, et al: Being 125 Gallagher S: Understanding interpersonal neural prediction problem. Neuron 2013;79: with virtual others: neural correlates of so- problems in autism: interaction theory as an 836–848. cial interaction. Neuropsychologia 2006; 44: alternative to theory of mind. Philos Psychi- 88 Heyes C: What’s social about social learn- 718–730. atry Psychol 2004;11:199–217. ing? J Comp Psychol 2012;126:193–202. 107 De Jaegher H, Di Paolo E: Participatory 126 Tom Froese: Overview. Dr. Tom Froese. 89 Behrens TEJ, Hunt LT, Woolrich MW, sense-making. Phenom Cogn Sci 2007; 6: https://froese.wordpress.com/research/ Rushworth MFS: Associative learning of so- 485–507. overview/ (last accessed September 9, 2017). cial value. Nature 2008;456:245–249. 108 Oullier O, de Guzman GC, Jantzen KJ, La- 127 Menary R: Introduction to the special issue 90 Apps MAJ, Green R, Ramnani N: Reinforce- garde J, Scott Kelso JA: Social coordination on 4E cognition. Phenom Cogn Sci 2010; 9: ment learning signals in the anterior cingu- dynamics: measuring human bonding. Soc 459–463. late cortex code for others’ false beliefs. Neu- Neurosci 2008;3:178–192. 128 Ward D, Stapleton M: Es are good: cognition roimage 2013;64:1–9. 109 Reddy V: How Infants Know Minds. Cam- as enacted, embodied, embedded, affective 91 Balsters JH, Apps MAJ, Bolis D, Lehner R, bridge, Harvard University Press, 2010. and extended; in Paglieri F (ed): Conscious- Gallagher L, Wenderoth N: Disrupted pre- 110 Hari R, Kujala MV: Brain basis of human ness in Interaction: The Role of the Natural diction errors index social deficits in autism social interaction: from concepts to brain and Social Environment in Shaping Con- spectrum disorder. Brain 2017;140:235–246. imaging. Physiol Rev 2009;89:453–479. sciousness. Advances in Consciousness Re- 92 Kilner JM: More than one pathway to action 111 De Jaegher H, Di Paolo E, Gallagher S: Can search. Amsterdam, Benjamins, 2012, vol understanding. Trends Cogn Sci 2011; 15: social interaction constitute social cogni- 86, pp 89–104. 352–357. tion? Trends Cogn Sci 2010;14:441–447. 129 Varela FJ, Thompson E, Rosch E: The Em- 93 Kilner JM, Friston KJ, Frith CD: Predictive 112 Schilbach L, Wilms M, Eickhoff SB, Roman- bodied Mind: Cognitive Science and Human coding: an account of the mirror neuron sys- zetti S, Tepest R, Bente G, et al: Minds made Experience. Cambridge, MIT Press, 1991. tem. Cogn Process 2007;8:159–166. for sharing: initiating joint attention recruits 130 Hurley SL: Consciousness in Action. Cam- 94 Hamilton AF, Grafton ST: The motor hier- reward-related neurocircuitry. J Cogn Neu- bridge, Harvard University Press, 1998. archy: from kinematics to goals and inten- rosci 2010;22:2702–2715. 131 Thompson E: Mind in Life: Biology, Phe- tions; in Haggard P, Rosetti Y, Kawato M 113 Froese T, Di Paolo EA: The enactive ap- nomenology, and the Sciences of Mind. (eds): Sensorimotor Foundations of Higher proach. Theoretical sketches from cell to so- Cambridge, Belknap Press of Harvard Uni- Cognition: Attention and Performance. Ox- ciety*. Pragmat Cogn 2011;19:1–36. versity Press, 2007. ford, Oxford University Press, 2007, pp 381– 114 Froese T, Fuchs T: The extended body: a case 132 Haugeland J: Having Thought: Essays in the 407. study in the neurophenomenology of social Metaphysics of Mind. Cambridge, Harvard 95 Becchio C, Manera V, Sartori L, Cavallo A, interaction. Phenom Cogn Sci 2012;11:205– University Press, 1998. Castiello U: Grasping intentions: from 235. 133 Clark A: Being There: Putting Brain, Body, thought experiments to empirical evidence. 115 Di Paolo E, De Jaegher H: The interactive and World Together Again. Cambridge, Front Hum Neurosci 2012;6:117. brain hypothesis. Front Hum Neurosci MIT Press, 1997. 96 Ansuini C, Cavallo A, Bertone C, Becchio C: 2012;6:163. 134 Gallagher S: How the Body Shapes the Mind. Intentions in the brain: the unveiling of 116 Konvalinka I, Roepstorff A: The two-brain Oxford, Oxford University Press, 2005. Mister Hyde. Neuroscientist 2015; 21: 126– approach: how can mutually interacting 135 Clark A, Chalmers D: The extended mind. 135. brains teach us something about social inter- Analysis 1998;58:7–19. 97 Huang Y, Rao RPN: Predictive coding. Wi- action? Front Hum Neurosci 2012;6:215. 136 Colombetti G: Enactive appraisal. Phenom ley Interdiscip Rev Cogn Sci 2011; 2: 580– 117 Dumas G: Towards a two-body neurosci- Cogn Sci 2007;6:527–546. 593. ence. Commun Integr Biol 2011;4:349–352. 137 Ratcliffe M: Existential feeling and psycho- 98 Rao RPN, Ballard DH: Predictive coding in 118 Dumas G, Martinerie J, Soussignan R, Nadel pathology. Philos Psychiatry Psychol 2009; the visual cortex: a functional interpretation J: Does the brain know who is at the origin 16:179-194. of some extra-classical receptive-field ef- of what in an imitative interaction? Front 138 Dumas G, de Guzman GC, Tognoli E, Kelso fects. Nat Neurosci 1999;2:79–87. Hum Neurosci 2012;6:128. JAS: The human dynamic clamp as a para- 99 Happé FG: Studying weak central coherence 119 Timmermans B, Schilbach L, Pasquali A, digm for social interaction. Proc Natl Acad at low levels: children with autism do not Cleeremans A: Higher order thoughts in Sci USA 2014;111:E3726–E3734. succumb to visual illusions. A research note. action: consciousness as an unconscious 139 Friston K, Frith C: A duet for one. Conscious J Child Psychol Psychiatry 1996;37:873–877. re-description process. Philos Trans R Soc Cogn 2015;36:390–405. 100 Kopp B: A simple hypothesis of executive Lond B Biol Sci 2012;367:1412–1423. 140 Friston KJ, Frith CD: Active inference, com- function. Front Hum Neurosci 2012;6:159. 120 Schilbach L, Timmermans B, Reddy V, Cos- munication and hermeneutics. Cortex 2015; 101 Miller GA: Plans and the Structure of Behav- tall A, Bente G, Schlicht T, et al: Toward a 68:129–143. ior. New York, Holt, 1960. second-person neuroscience. Behav Brain 141 Vittorias J, Petrantonakis P, Bolis D, Tsilig- 102 Badre D, D’Esposito M: Functional magnet- Sci 2013;36:393–414. kyri A, Kosmidou V, Hadjileontiadis LJ: ic resonance imaging evidence for a hierar- 121 Dale R, Fusaroli R, Duran ND, Richardson NOESIS: an enhanced educational environ- chical organization of the prefrontal cortex. DC: The self-organization of human inter- ment for kids with autism spectrum disor- J Cogn Neurosci 2007;19:2082–2099. action. Psychol Learn Motiv Adv Res Theory ders. 8th IEEE International Conference on 103 Badre D, D’Esposito M: Is the rostro-caudal 2013;59:43–95. Advanced Learning Technologies, Santan- axis of the frontal lobe hierarchical? Nat Rev 122 Schilbach L: On the relationship of online dar, 2008, pp 1019–1020. Neurosci 2009;10:659–669. and offline social cognition. Front Hum 142 Laing RD, Phillipson H, Lee AR: Interper- 104 Quattrocki E, Friston K: Autism, oxytocin Neurosci 2014;8:278. sonal perception; a theory and a method of and interoception. Neurosci Biobehav Rev 123 Schilbach L: Towards a second-person neu- research. New York, Harper & Row, 1972. 2014;47:410–430. ropsychiatry. Philos Trans R Soc London B Biol Sci 2015;371:20150081. 193.175.245.253 - 1/11/2018 5:02:03 PM The Dialectical Misattunement Psychopathology 2017;50:355–372 371 Hypothesis DOI: 10.1159/000484353 Max Planck Society Downloaded by: 143 Mathys C, Daunizeau J, Friston KJ, Stephan 147 Friston K: Life as we know it. J R Soc Inter- 151 Adorno TW: An Introduction to Dialectics. KE: A Bayesian foundation for individual face 2013;10:20130475. Hoboken, John Wiley & Sons, 1958/2017. learning under uncertainty. Front Hum 148 Kelso JAS: Dynamic Patterns: The Self-Or- 152 Robinson H: Dualism; in Zalta EN (ed): The Neurosci 2011;5:39. ganization of Brain and Behavior. Cam- Stanford Encyclopedia of Philosophy (Fall 144 Bolis D, Heinzle J, Mathys C, Stephan KE: bridge, MIT Press, 1995. 2017 Edition). https://plato.stanford.edu/ Inferring Cognitive Traits of Individual 149 Hirsch ED, Kett JF, Trefil JS: The New Dic- archives/fall2017/entries/dualism/. Subjects through Gaze Controlled Video tionary of Cultural Literacy. Boston, Hough- 153 O’Connor T, Wong HY: Emergent Proper- Games. Trieste, EUT Edizioni Università di ton Mifflin Harcourt, 2002. ties; in Zalta EN (ed): The Stanford Encyclo- Trieste, 2015. 150 Toomela A: There can be no cultural-histor- pedia of Philosophy (Summer 2015 Edition). 145 Pessoa L: On the relationship between emo- ical psychology without neuropsychology. https://plato.stanford.edu/archives/sum tion and cognition. Nat Rev Neurosci 2008; And vice versa; in Yasnitsky A, van der Veer 2015/entries/properties-emergent/. 9:148–158. R, Ferrari M (eds): The Cambridge Hand- 154 Haken H: Synergetics: an overview. Rep 146 Stefanics G, Kremláček J, Czigler I: Visual book of Cultural-Historical Psychology. Prog Phys 1989;52:515. mismatch negativity: a predictive coding Cambridge, Cambridge University Press, view. Front Hum Neurosci 2014;8:666. 2014, pp 313–349. 193.175.245.253 - 1/11/2018 5:02:03 PM 372 Psychopathology 2017;50:355–372 Bolis/Balsters/Wenderoth/Becchio/ DOI: 10.1159/000484353 Schilbach Max Planck Society Downloaded by: