Rett Syndrome is a rare neurodevelopmental disorder associated with severe intellectual and motor impairments. Among its core symptoms, children with Rett progressively lose purposeful hand use, limiting their ability to reach for, touch, and manipulate objects in the surrounding environment. Early intervention is therefore important to help maintain opportunities for functional hand use through repeated and motivating activities. To support this goal, we developed Pufff, a touch-sensitive interactive toy that provides visual feedback during touch-based interaction. Using signals acquired from a six-patch capacitive sensing array and a lightweight machine learning model, the system distinguishes three types of touch and triggers different responses accordingly. This interaction paradigm is intended to support future therapist-led activities aimed at encouraging purposeful hand use through play, while also enabling basic monitoring of touch interaction patterns. This paper presents an early implementation of the toy and demonstrates the technical feasibility of the approach.

Figure 1: A screenshot of the turn-taking play using the interactive soft toy Octopus X-8.The data collected by the toy, combined with the data collected by the smart glasses worn by the therapist for eye contact detection, provide an objective evaluation of the child's social interaction with the caregiver during the play activities (child's face is shown with parental consent).

In this paper, we provide an in-depth description of Echo, an interactive, smart toy, designed to encourage the symbolic play in autistic children. Symbolic play is a key competence in the cognitive development, and it is observed for instance when a child uses an object to represent other objects. The toy, characterised by an abstract shape, can autonomously recognise how the user handles it during play (e.g. mimicking the movements of a car, an aeroplane, a frog in this study) and produce, as rewarding feedback, distinctive sensory outputs such as colours and sounds, according to the detected imaginary toy. The categorisation, achieved with an average accuracy of 90.1%, is carried out in real-time by a Machine Learning (ML) model, which runs in the device's embedded electronics; the ML model was trained using the data by an onboard IMU sensor, which provides information about the toy movements (i.e. tilt, angular velocity, acceleration). The paper provides an overview of the toy design and functionalities, and presents the results of a feasibility test with neurotypical adult participants. Finally, it proposes Echo as a potential tool to encourage, through the AI-mediated sensory feedback, this pivotal competence; symbolic play is in fact often impaired or atypical in autistic children, and new technologies – such as Echo – can provide interesting novel opportunities for neurodevelopmental therapists and researchers.

Even in the absence of external stimuli, the brain is spontaneously active. Indeed, most cortical activity is internally generated by recurrence. Both theoretical and experimental studies suggest that chaotic dynamics characterize this spontaneous activity. While the precise function of brain chaotic activity is still puzzling, we know that chaos confers many advantages. From a computational perspective, chaos enhances the complexity of network dynamics. From a behavioural point of view, chaotic activity could generate the variability required for exploration. Furthermore, information storage and transfer are maximized at the critical border between order and chaos. Despite these benefits, many computational brain models avoid incorporating spontaneous chaotic activity due to the challenges it poses for learning algorithms. In recent years, however, multiple approaches have been proposed to overcome this limitation. As a result, many different algorithms have been developed, initially within the reservoir computing paradigm. Over time, the field has evolved to increase the biological plausibility and performance of the algorithms, sometimes going beyond the reservoir computing framework. In this review article, we examine the computational benefits of chaos and the unique properties of chaotic recurrent neural networks, with a particular focus on those typically utilized in reservoir computing. We also provide a detailed analysis of the algorithms designed to train chaotic RNNs, tracing their historical evolution and highlighting key milestones in their development. Finally, we explore the applications and limitations of chaotic RNNs for brain modelling, consider their potential broader impacts beyond neuroscience, and outline promising directions for future research.

The role of the dorsal anterior cingulate cortex (dACC) in cognition is a frequently studied yet highly debated topic in neuroscience. Most authors agree that the dACC is involved in either cognitive control (e.g., voluntary inhibition of automatic responses) or monitoring (e.g., comparing expectations with outcomes, detecting errors, tracking surprise). A consensus on which theoretical perspective best explains dACC contribution to behaviour is still lacking, as two distinct sets of studies report dACC activation in tasks requiring surprise tracking for performance monitoring and cognitive control without involving surprise monitoring, respectively. This creates a theoretical impasse, as no single current account can reconcile these findings. Here we propose a novel hypothesis on dACC function that integrates both the monitoring and the cognitive control perspectives in a unifying, meta-Reinforcement Learning framework, in which cognitive control is optimized by meta-learning based on tracking Bayesian surprise. We tested the quantitative predictions from our theory in three different functional neuroimaging experiments at the basis of the current theory crisis. We show that the meta-Reinforcement Learning perspective successfully captures all the neuroimaging results by predicting both cognitive control and monitoring functions, proposing a solution to the theory crisis about dACC function within an integrative framework. In sum, our results suggest that dACC function can be framed as a meta-learning optimisation of cognitive control, providing an integrative perspective on its roles in cognitive control, surprise tracking, and performance monitoring.

Emotion-Lab is a novel educational model for the development of Emotional Intelligence in children. It is based on the use of innovative, interactive smart toys, called Transitional Wearable Companions (TWCs), originally designed to stimulate social engagement in children with Autism Spectrum Disorders (ASD), through sensory-motor play with the therapist. In this work the TWCs are used in a new, educational scenario addressed to typically developed children. In a pilot test, teachers and pedagogists used the interactive features of TWCs to set up learning activities which aim to develop and improve emotional awareness and emotional competence, fundamental skills in a child’s cognitive development. In more detail the Emotion-Lab, through inclusive, social activities, aims to encourage participants to (a) become more aware of their emotional states; (b) learn to distinguish and understand the emotions of others; (c) express their own emotions appropriately; (d) experience empathic and sympathetic involvement; (e) distinguish between manifest emotional states and actual internal states. At the end of the test, children were then encouraged, through discussion, to share their intra and interpersonal experiences, with the main objective of improving mutual understanding and increasing social cohesion and, at the same time, reducing the phenomena of social discomfort, bullying, and cyberbullying.

This paper describes a tangible, interactive platform presented here as an early prototype, designed to evaluate the child’s social interaction with the caregiver, during play activities based on turn-taking rules. The platform is composed by 2 elements, each one providing relevant behavioural data: (a) an interactive soft toy, called Octopus X-8, able to discriminate between 2 users and to respond accordingly, by emitting rewarding sensory feedback when the game rules are observed: this toy provides data about the turn-taking interaction (e.g. users’ observance of the turn, timing of rewards, sensors activity); (b) a pair of smart glasses worn by the caregiver, which provides an AI-based evaluation of the eye contact with the child. At the end of the play session, the data from the 2 sources are merged, so as to produce an overall, objective evaluation of the social interaction between the two users. A pilot study, involving a child undergoing neuropsychological assessment, shows how collected data are coherent with the video recording of the experimental session. We then propose the platform as a potential tool to monitor the behaviour of children with neurodevelopmental disorders.

Goal-directed manipulation of internal representations is a key element of human flexible behaviour, while consciousness is commonly associated with higher-order cognition and human flexibility. Current perspectives have only partially linked these processes, thus preventing a clear understanding of how they jointly generate flexible cognition and behaviour. Moreover, these limitations prevent an effective exploitation of this knowledge for technological scopes. We propose a new theoretical perspective that extends our 'three-component theory of flexible cognition' toward higher-order cognition and consciousness, based on the systematic integration of key concepts from Cognitive Neuroscience and AI/Robotics. The theory proposes that the function of conscious processes is to support the alignment of representations with multi-level goals. This higher alignment leads to more flexible and effective behaviours. We analyse here our previous model of goal-directed flexible cognition (validated with more than 20 human populations) as a starting GARIM-inspired model. By bridging the main theories of consciousness and goal-directed behaviour, the theory has relevant implications for scientific and technological fields. In particular, it contributes to developing new experimental tasks and interpreting clinical evidence. Finally, it indicates directions for improving machine learning and robotics systems and for informing real-world applications (e.g., in digital-twin healthcare and roboethics).

This paper presents the preliminary results where an experimental prototype, the interactive soft toy called Octopus X-8, is used with three children with neurodevelopmental disorders in play activities involving turn-taking. This competence is fundamental for regulating social interactions and it is often impaired in those types of disorders. The pilot experiment aims to investigate if X-8 can be used as an engaging toy usable to train turn-taking skills. The toy is able to distinguish between two users and can therefore produce rewarding sensory feedback, such as coloured lights and sounds, when the turn-taking rules are respected. Preliminary results seem to show that X-8 can indeed be used as a supporting tool to improve this important social competence.

In this workshop we will present, through live demos, novel technological tools such as interactive "smart" toys for social play and AI-based monitoring tools, developed to support the early treatment of Autism Spectrum Disorder (ASD). We will discuss the potential of such devices, describing also our promising results with ASD children. The goal of the workshop is to involve the specialist audience in the co-design of some potential aspects of the overall system. In particular, we are interested in collecting feedback and proposals on possible further uses of the system such as additional play activities to stimulate social behaviour; improvements of the monitoring tools; potential use of the tools in developmental disorders other than ASD.

As human beings, we are designed to be empathic, to connect emotionally with each other and to share affection. This is part of our nature and makes us feel like part of a group. New forms of technology enable us to share, in real time, a variety of data. Some data, such as those on heart rate, can relate to our affective state. This research focuses on asking if technology-mediated interactions might help us to feel more emotionally engaged and intimately connected, especially when sensorial feedback about our inner state is provided. In this context, we present an interactive device that provides two users, through pulsing lights, with feedback on their heartbeats and the related level of synchrony (fig. 1). The prototype could be used to assess the effectiveness of this technology and to improve the feeling of connectedness and intimacy between two users.

Transitional Wearable Companion (TWC) is a novel design concept, implemented as an interactive, smart, soft toy, which aims to stimulate curiosity and encourage social engagement in kids with special needs, during play activities with their caregivers. We propose the TWC as a potential support tool for neurodevelopmental therapists, during early intervention in disorders characterised by impairment in the social area, such as Autism Spectrum Disorder (ASD). The TWC might be helpful to set up sensory-motor games, conceived for encouraging and reinforcing critical social competences, pivotal for cognitive development, such as eye-contact, imitation, joint-attention and turn-taking. In this work, we present two working prototypes of TWC called PlusMe and X-8, currently used in pilot experiments with children diagnosed with ASD.

Executive functions represent a set of processes in goal-directed cognition that depend on integrated cortical-basal ganglia brain systems and form the basis of flexible human behaviour. Several computational models have been proposed for studying cognitive flexibility as a key executive function and the Wisconsin card sorting test (WCST) that represents an important neuropsychological tool to investigate it. These models clarify important aspects that underlie cognitive flexibility, particularly decision-making, motor response, and feedback-dependent learning processes. However, several studies suggest that the categorisation processes involved in the solution of the WCST include an additional computational stage of category representation that supports the other processes. Surprisingly, all models of the WCST ignore this fundamental stage and they assume that decision making directly triggers actions. Thus, we propose a novel hypothesis where the key mechanisms of cognitive flexibility and goal-directed behaviour rely on the acquisition of suitable representations of percepts and their top-down internal manipulation. Moreover, we propose a neuro-inspired computational model to operationalise this hypothesis. The capacity of the model to support cognitive flexibility was validated by systematically reproducing and interpreting the behaviour exhibited in the WCST by young and old healthy adults, and by frontal and Parkinson patients. The results corroborate and further articulate the hypothesis that the internal manipulation of representations is a core process in goal-directed flexible cognition.

This paper presents the achievements obtained from a study performed within the IMPACT (Intrinsically Motivated Planning Architecture for Curiosity-driven roboTs) Project funded by the European Space Agency (ESA).The main contribution of the work is the realization of an innovative robotic architecture in which the well-known three-layered architectural paradigm (decisional, executive, and functional) for controlling robotic systems is enhanced with autonomous learning capabilities.The architecture is the outcome of the application of an interdisciplinary approach integrating Artificial Intelligence (AI), Autonomous Robotics, and Machine Learning (ML) techniques.In particular, state-of-the-art AI planning systems and algorithms were integrated with Reinforcement Learning (RL) algorithms guided by intrinsic motivations (curiosity, exploration, novelty, and surprise).The aim of this integration was to: (i) develop a software system that allows a robotic platform to autonomously represent in symbolic form the skills autonomously learned through intrinsic motivations; (ii) show that the symbolic representation can be profitably used for automated planning purposes, thus improving the robot's exploration and knowledge acquisition capabilities.The proposed solution is validated in a test scenario inspired by a typical space exploration mission involving a rover.

In mammals, goal-directed and planning processes support flexible behaviour used to face new situations that cannot be tackled through more efficient but rigid habitual behaviours. Within the Bayesian modelling approach of brain and behaviour, models have been proposed to perform planning as probabilistic inference but this approach encounters a crucial problem: explaining how such inference might be implemented in brain spiking networks. Recently, the literature has proposed some models that face this problem through recurrent spiking neural networks able to internally simulate state trajectories, the core function at the basis of planning. However, the proposed models have relevant limitations that make them biologically implausible, namely their world model is trained 'off-line' before solving the target tasks, and they are trained with supervised learning procedures that are biologically and ecologically not plausible. Here we propose two novel hypotheses on how brain might overcome these problems, and operationalise them in a novel architecture pivoting on a spiking recurrent neural network. The first hypothesis allows the architecture to learn the world model in parallel with its use for planning: to this purpose, a new arbitration mechanism decides when to explore, for learning the world model, or when to exploit it, for planning, based on the entropy of the world model itself. The second hypothesis allows the architecture to use an unsupervised learning process to learn the world model by observing the effects of actions. The architecture is validated by reproducing and accounting for the learning profiles and reaction times of human participants learning to solve a visuomotor learning task that is new for them. Overall, the architecture represents the first instance of a model bridging probabilistic planning and spiking-processes that has a degree of autonomy analogous to the one of real organisms.

Learning in biologically relevant neural-network models usually relies on Hebb learning rules. The typical implementations of these rules change the synaptic strength on the basis of the co-occurrence of the neural events taking place at a certain time in the pre- and post-synaptic neurons. Differential Hebbian learning (DHL) rules, instead, are able to update the synapse by taking into account the temporal relation, captured with derivatives, between the neural events happening in the recent past. The few DHL rules proposed so far can update the synaptic weights only in few ways: this is a limitation for the study of dynamical neurons and neural-network models. Moreover, empirical evidence on brain spike-timing-dependent plasticity (STDP) shows that different neurons express a surprisingly rich repertoire of different learning processes going far beyond existing DHL rules. This opens up a second problem of how capturing such processes with DHL rules. Here we propose a general DHL (G-DHL) rule generating the existing rules and many others. The rule has a high expressiveness as it combines in different ways the pre- and post-synaptic neuron signals and derivatives. The rule flexibility is shown by applying it to various signals of artificial neurons and by fitting several different STDP experimental data sets. To these purposes, we propose techniques to pre-process the neural signals and capture the temporal relations between the neural events of interest. We also propose a procedure to automatically identify the rule components and parameters that best fit different STDP data sets, and show how the identified components might be used to heuristically guide the search of the biophysical mechanisms underlying STDP. Overall, the results show that the G-DHL rule represents a useful means to study time-sensitive learning processes in both artificial neural networks and brain.

Motor tics are a cardinal feature of Tourette syndrome and are traditionally associated with an excess of striatal dopamine in the basal ganglia. Recent evidence increasingly supports a more articulated view where cerebellum and cortex, working closely in concert with basal ganglia, are also involved in tic production. Building on such evidence, this article proposes a computational model of the basal ganglia-cerebellar-thalamo-cortical system to study how motor tics are generated in Tourette syndrome. In particular, the model: (i) reproduces the main results of recent experiments about the involvement of the basal ganglia-cerebellar-thalamo-cortical system in tic generation; (ii) suggests an explanation of the system-level mechanisms underlying motor tic production: in this respect, the model predicts that the interplay between dopaminergic signal and cortical activity contributes to triggering the tic event and that the recently discovered basal ganglia-cerebellar anatomical pathway may support the involvement of the cerebellum in tic production; (iii) furnishes predictions on the amount of tics generated when striatal dopamine increases and when the cortex is externally stimulated. These predictions could be important in identifying new brain target areas for future therapies. Finally, the model represents the first computational attempt to study the role of the recently discovered basal ganglia-cerebellar anatomical links. Studying this non-cortex-mediated basal ganglia-cerebellar interaction could radically change our perspective about how these areas interact with each other and with the cortex. Overall, the model also shows the utility of casting Tourette syndrome within a system-level perspective rather than viewing it as related to the dysfunction of a single brain area.

Mental rotation, a classic experimental paradigm of cognitive psychology, tests the capacity of humans to mentally rotate a seen object to decide if it matches a target object. In recent years, mental rotation has been investigated with brain imaging techniques to identify the brain areas involved. Mental rotation has also been investigated through the development of neural-network models, used to identify the specific mechanisms that underlie its process, and with neurorobotics models to investigate its embodied nature. Current models, however, have limited capacities to relate to neuro-scientific evidence, to generalise mental rotation to new objects, to suitably represent decision making mechanisms, and to allow the study of the effects of overt gestures on mental rotation. The work presented in this study overcomes these limitations by proposing a novel neurorobotic model that has a macro-architecture constrained by knowledge held on brain, encompasses a rather general mental rotation mechanism, and incorporates a biologically plausible decision making mechanism. The model was tested using the humanoid robot iCub in tasks requiring the robot to mentally rotate 2D geometrical images appearing on a computer screen. The results show that the robot gained an enhanced capacity to generalise mental rotation to new objects and to express the possible effects of overt movements of the wrist on mental rotation. The model also represents a further step in the identification of the embodied neural mechanisms that may underlie mental rotation in humans and might also give hints to enhance robots' planning capabilities.

We demonstrate a sample-efficient method for constructing reusable parameterized skills that can solve families of related motor tasks. Our method uses learned policies to analyze the policy space topology and learn a set of regression models which, given a novel task, appropriately parameterizes an underlying low-level controller. By identifying the disjoint charts that compose the policy manifold, the method can separately model the qualitatively different sub-skills required for solving distinct classes of tasks. Such sub-skills are useful because they can be treated as new discrete, specialized actions by higher-level planning processes. We also propose a method for reusing seemingly unsuccessful policies as additional, valid training samples for synthesizing the skill, thus accelerating learning. We evaluate our method on a humanoid iCub robot tasked with learning to accurately throw plastic balls at parameterized target locations.

Organisms that live in groups, from microbial symbionts to social insects and schooling fish, exhibit a number of highly efficient cooperative behaviors, often based on role taking and specialization. These behaviors are relevant not only for the biologist but also for the engineer interested in decentralized collective robotics. We address these phenomena by carrying out experiments with groups of two simulated robots controlled by neural networks whose connection weights are evolved by using genetic algorithms. These algorithms and controllers are well suited to autonomously find solutions for decentralized collective robotic tasks based on principles of self-organization. The article first presents a taxonomy of role-taking and specialization mechanisms related to evolved neural network controllers. Then it introduces two cooperation tasks, which can be accomplished by either role taking or specialization, and uses these tasks to compare four different genetic algorithms to evaluate their capacity to evolve a suitable behavioral strategy, which depends on the task demands. Interestingly, only one of the four algorithms, which appears to have more biological plausibility, is capable of evolving role taking or specialization when they are needed. The results are relevant for both collective robotics and biology, as they can provide useful hints on the different processes that can lead to the emergence of specialization in robots and organisms.

Despite the great amount of knowledge produced by the neuroscientific literature on affective phenomena, current models tackling non-cognitive aspects of behaviour are often bio-inspired but rarely bio-constrained. This paper presents a theoretical account of affective systems centred on the amygdala (Amg). This account aims to furnish a general framework and specific pathways to implement models that are more closely related to biological evidence. The Amg, which receives input from brain areas encoding internal states, innately relevant stimuli, and innately neutral stimuli, plays a fundamental role in the motivational and emotional processes of organisms. This role is based on the fact that Amg implements the two associative processes at the core of Pavlovian learning (conditioned stimulus (CS)–unconditioned stimulus (US) and CS–unconditioned response (UR) associations), and that it has the capacity of modulating these associations on the basis of internal states. These functionalities allow the Amg to play an important role in the regulation of the three fundamental classes of affective responses (namely, the regulation of body states, the regulation of brain states via neuromodulators, and the triggering of a number of basic behaviours fundamental for adaptation) and in the regulation of three high-level cognitive processes (namely, the affective labelling of memories, the production of goal-directed behaviours, and the performance of planning and complex decision-making). Our analysis is conducted within a methodological approach that stresses the importance of understanding the brain within an evolutionary/adaptive framework and with the aim of isolating general principles that can potentially account for the wider possible empirical evidence in a coherent fashion.

This paper analyses the strengths and weaknesses of self-organising approaches, such as evolutionary ro-botics, and direct design approaches, such as behav-iour-based controllers, for the production of autono-mous robots? controllers, and shows how the two approaches can be usefully combined. In particular, the paper proposes a method for encoding evolved neural-network based behaviours into motor schema-based controllers and then shows how these control-lers can be modified and combined to produce robots capable of solving new tasks. The method has been validated in the context of a collective robotics sce-nario in which a group of physically assembled simulated autonomous robots are requested to pro-duce different forms of coordinated behaviours (e.g., coordinated motion, walled-arena exiting, and light pursuing).

An important goal of collective robotics is the design of control systems that allow groups of robots to accomplish common tasks by coordinating without a centralized control. In this paper, we study how a group of physically assembled robots can display coherent behavior on the basis of a simple neural controller that has access only to local sensory information. This controller is synthesized through artificial evolution in a simulated environment in order to let the robots display coordinated-motion behaviors. The evolved controller proves to be robust enough to allow a smooth transfer from simulated to real robots. Additionally, it generalizes to new experimental conditions, such as different sizes/shapes of the group and/or different connection mechanisms. In all these conditions the performance of the neural controller in real robots is comparable to the one obtained in simulation.

Distributed coordination of groups of individuals accomplishing a common task without leaders, with little communication, and on the basis of self-organizing principles, is an important research issue within the study of collective behavior of animals, humans, and robots. The article shows how distributed coordination allows a group of evolved, physically linked simulated robots (inspired by a robot under construction) to display a variety of highly coordinated basic behaviors such as collective motion, collective obstacle avoidance, and collective approach to light, and to integrate them in a coherent fashion. In this way the group is capable of searching and approaching a lighted target in an environment scattered with obstacles, furrows, and holes, where robots acting individually fail. The article shows how the emerged coordination of the group relies upon robust self-organizing principles (e.g., positive feedback) based on a novel sensor that allows the single robots to perceive the group's "average" motion direction. The article also presents a robust solution to a difficult coordination problem, which might also be encountered by some organisms, caused by the fact that the robots have to be capable of moving in any direction while being physically connected. Finally, the article shows how the evolved distributed coordination mechanisms scale very well with respect to the number of robots, the way in which robots are assembled, the structure of the environment, and several other aspects.

We present a set of experiments in which simulated robots are evolved for the ability to aggregate and move together toward a light target. By developing and using quantitative indexes that capture the structural properties of the emerged formations, we show that evolved individuals display interesting behavioral patterns in which groups of robots act as a single unit. Moreover, evolved groups of robots with identical controllers display primitive forms of situated specialization and play different behavioral functions within the group according to the circumstances. Overall, the results presented in the article demonstrate that evolutionary techniques, by exploiting the self-organizing behavioral properties that emerge from the interactions between the robots and between the robots and the environment, are a powerful method for synthesizing collective behavior.

Is it possible to plan at a coarse level and act at a fine level with a neural-network (NN) reinforcement-learning (RL) planner? This work presents a NN planner, used to control a simulated robot in a stochastic landmark-navigation problem, which plans at an abstract level. The controller has both reactive components, based on actor-critic RL, and planning components inspired by the Dyna-PI architecture (this roughly corresponds to RL plus a model of the environment). Coarse planning is based on macro-actions defined as a sequence of identical primitive actions. It updates the evaluations and the action policy while generating simulated experience at the macro level with the model of the environment (a NN trained at the macro level). The simulations show how the controller works. They also show the advantages of using a discount coefficient tuned to the level of planning coarseness, and suggest that discounted RL has problems in dealing with long periods of time.