That psychological sciences suffer from a profound crisis is probably not extremely controversial. Yet, arguably, the recently debated replication failure is nothing but a symptom of deeply rooted dichotomies and ontological commitments lying at its core. Undeniably, essential aspects of the human condition are typically studied in isolation via applying static categories, while philosophical considerations and human practice are largely neglected. In this context and drawing inspiration from real-life experience through a Vygotskian lens, this thesis attempts to motivate a systematic shift of focus from being to becoming; in fact becoming-with. More concretely, leaning on the dialectical method, cultural-historical theory and recent developments of social computational neuroscience, (i.) this thesis presents the dialectical attunement account which argues that a multiscale analysis of social interaction might allow us to scientifically reconsider the self, beyond the individual, where it really emerges, unfolds and manifests itself — in social relations. In this light, (ii.) it puts forward the dialectical misattunement hypothesis, which views autism and broadly psychopathology as a dynamic interpersonal mismatch, rather than a (disordered) function of single brains. Critically, (iii.) it operationalizes these hypotheses by establishing a novel empirical framework, namely two-person psychophysiology, which measures and analyzes the multiscale dynamics of social interaction. Deploying this framework, this thesis empirically demonstrates that (iv.) real-time dynamics of social interaction do matter in both collective and individual dimensions ‒even beyond awareness‒ lending support to second-person and enactivist proposals. With regard to psychopathology, this thesis demonstrates that (v.) it is primarily the mismatch of autistic traits –not traits per se– which predicts core aspects of interpersonal attunement in real-life social relations, offering a first empirical validation of the dialectical misattunement hypothesis. Taken together, this thesis tries to break free from dichotomies such as internalism/externalism or healthy/patient, in experiential, theoretical, methodological and empirical regards. Such a dialectical and empirical approach to human becoming in and through social interaction encourages a social change pertinent to various fields of human research and practice, ranging from psychiatry and pedagogy to ethics and artificial intelligence.

This is a theoretical effort of bringing together seemingly diverse approaches for gaining a better insight in the co-construction of social and individual reality. So far, it has focused on rethinking psychiatric and psychological definitions with an emphasis on autism. Leaning on sociocultural-historical activity theories, enactivism, social neuroscience, and Bayesian accounts, the first step has been an attempt on reconciling dialectical and mechanistic views on life, social-interaction, the mind and culture.

Here, we have developed a framework for studying multilevel mechanisms, i.e. from intra- to interpersonal processes, as well as their interrelationships in individuals with and without psychiatric conditions. Due to conceptual and methodological constraints, neuropsychiatric research has focused largely on (mal-)adaptive mechanisms at the individual level. Our setup allows for real-time interactions between participants through optical micro-cameras while also obtaining high-resolution empirical data via infrared eye-trackers and biological motion sensors, thus, providing multiple behavioral readouts, such as facial expression and gaze position.

In this project resting state fMRI data of a large sample of population from multiple research centers is being analyzed. By employing advanced computational methods we are aiming at investigating both inter- and intra-group neural differences and connectivity between people having received a diagnosis of Autism Spectrum Condition (ASC), people with a diagnosis of Attention Deficit Hyperactivity Disorder (ADHD) and individuals without such a diagnosis.

In this project, we investigated the efficiency of an approach that combines video games, eye movement tracking and Bayesian modeling for studying learning and decision-making in humans. Crucially, two ways of interaction (i.e. eye or hand) with the environment were offered and two probabilistic conditions (volatile or stable) were considered in two spatial dimensions. Subject's responses were monitored in two series of experiments. Learning was modeled based on classical conditioning theories and novel predictive coding schemes ( i.e. Hierarchical Gaussian Filtering). Additionally, a number of hypotheses concerning the translation from beliefs to responses (e.g. the relation between the confidence of a prediction and the response variability) were studied. Importantly, the experimental task allowed comparison of different controlling effectors, intra-effector goals and probabilistic condition impact, with an ultimate goal of studying certain spectrum conditions and particularly autism spectrum conditions in children through the lens of predictive coding.

In this project, we hypothesized that thalamocortical connections after mapping by diffusion tensor imaging can serve as surrogate markers of individual anatomy, which can then be used for localizing specific neurosurgical targets in the thalamus. A variety of learning schemes, together with pre- and post- processing steps were studied for our goal of thalamic nuclei localization. The training procedure was performed after non-linear registration and probabilistic tractography on diffusion magnetic resonance imaging data. Our results indicated that thalamocortical connectivity data do contain sufficient discriminant internucleus information for thalamic nuclei localization.

The aim of the papers emerged out of this work has been three-fold. Firstly, to illustrate the sensitivity of subspace learning methods when the registration of the facial region of interest prior to recognition fails, even slightly. Moreover, to illustrate that the inter-database recognition performance is much worse than the intra-database performance that is usually reported in the literature. Secondly, to propose a training set enrichment approach for improving significantly the performance of subspace learning techniques in the facial expression recognition problem. Moreover, to highlight that even perfect manual face alignment in high resolution can be improved by the proposed training set enrichment. Thirdly, to indicate the contribution of enriching the training set with images of a tested person, in order to create person specific recognizers, thus, improving the subspace learning and the recognition performance.

This software tool is structured on a 3-level basis in order to optimize overall precision and speed. Firstly, a fully automatic, histogram based, algorithm is used for the scull and brain segmentation. In cases of extremely noisy slices, semi-automatic and manual mode are offered. The 3D brain model is reconstructed (modes of both volume and surface rendering are offered) and the brain's volume is calculated. Specific regions of interest (e.g. brain tumors) can be interactively selected and visualized together with the brain. Crucially, a user-friendly graphical user interface was designed and implemented for enabling non-specialists to effortlessly exploit the tool.

A novel educational environment for children with an autism spectrum condition (ASC), namely NOESIS, was developed during this project. NOESIS takes into account ASC child's individual characteristics, their emotional state during their educational procedure, and creativity during guided- and self-activity (e.g. gaming) and it importantly adapts to each child's specific characteristics through system adaptation and self-regulation procedures. Moreover, it provides assistance to the educator for preparation, customization and optimization of the educational material for each kid and provision of enhanced evaluation procedures (scores/tools) via well-managed Web services. Parents' updating is also provided via reporting material with learning curve descriptions. Overall, NOESIS contributes to the provision of opportunities to all ASC children to be educated by facilitating access and tuning innovative technology to social needs. In brief NOESIS views autism not as a disorder, but as a way of being.