How to train a self-driving vehicle: On the added value (or lack thereof) of curriculum learning and replay buffers.

Learning from only real-world collected data can be unrealistic and time consuming in many scenario. One alternative is to use synthetic data as learning environments to learn rare situations and replay buffers to speed up the learning. In this work, we examine the hypothesis of how the creation of the environment affects the training of reinforcement learning agent through auto-generated environment mechanisms. We take the autonomous vehicle as an application. We compare the effect of two approaches to generate training data for artificial cognitive agents. We consider the added value of curriculum learning-just as in human learning-as a way to structure novel training data that the agent has not seen before as well as that of using a replay buffer to train further on data the agent has seen before. In other words, the focus of this paper is on characteristics of the training data rather than on learning algorithms. We therefore use two tasks that are commonly trained early on in autonomous vehicle research: lane keeping and pedestrian avoidance. Our main results show that curriculum learning indeed offers an additional benefit over a vanilla reinforcement learning approach (using Deep-Q Learning), but the replay buffer actually has a detrimental effect in most (but not all) combinations of data generation approaches we considered here. The benefit of curriculum learning does depend on the existence of a well-defined difficulty metric with which various training scenarios can be ordered. In the lane-keeping task, we can define it as a function of the curvature of the road, in which the steeper and more occurring curves on the road, the more difficult it gets. Defining such a difficulty metric in other scenarios is not always trivial. In general, the results of this paper emphasize both the importance of considering data characterization, such as curriculum learning, and the importance of defining an appropriate metric for the task.

Reflection Machines: Supporting Effective Human Oversight Over Medical Decision Support Systems.

Human decisions are increasingly supported by decision support systems (DSS). Humans are required to remain "on the loop," by monitoring and approving/rejecting machine recommendations. However, use of DSS can lead to overreliance on machines, reducing human oversight. This paper proposes "reflection machines" (RM) to increase meaningful human control. An RM provides a medical expert not with suggestions for a decision, but with questions that stimulate reflection about decisions. It can refer to data points or suggest counterarguments that are less compatible with the planned decision. RMs think against the proposed decision in order to increase human resistance against automation complacency. Building on preliminary research, this paper will (1) make a case for deriving a set of design requirements for RMs from EU regulations, (2) suggest a way how RMs could support decision-making, (3) describe the possibility of how a prototype of an RM could apply to the medical domain of chronic low back pain, and (4) highlight the importance of exploring an RM's functionality and the experiences of users working with it.

The world seems different in a social context: A neural network analysis of human experimental data.

Human perception and behavior are affected by the situational context, in particular during social interactions. A recent study demonstrated that humans perceive visual stimuli differently depending on whether they do the task by themselves or together with a robot. Specifically, it was found that the central tendency effect is stronger in social than in non-social task settings. The particular nature of such behavioral changes induced by social interaction, and their underlying cognitive processes in the human brain are, however, still not well understood. In this paper, we address this question by training an artificial neural network inspired by the predictive coding theory on the above behavioral data set. Using this computational model, we investigate whether the change in behavior that was caused by the situational context in the human experiment could be explained by continuous modifications of a parameter expressing how strongly sensory and prior information affect perception. We demonstrate that it is possible to replicate human behavioral data in both individual and social task settings by modifying the precision of prior and sensory signals, indicating that social and non-social task settings might in fact exist on a continuum. At the same time, an analysis of the neural activation traces of the trained networks provides evidence that information is coded in fundamentally different ways in the network in the individual and in the social conditions. Our results emphasize the importance of computational replications of behavioral data for generating hypotheses on the underlying cognitive mechanisms of shared perception and may provide inspiration for follow-up studies in the field of neuroscience.

Implementing Affordable Socially Assistive Pet Robots in Care Homes Before and During the COVID-19 Pandemic: Stratified Cluster Randomized Controlled Trial and Mixed Methods Study.

BACKGROUND: Robot pets may assist in the challenges of supporting an aging population with growing dementia prevalence. Prior work has focused on the impacts of the robot seal Paro on older adult well-being, but recent studies have suggested the good acceptability and implementation feasibility of more affordable devices (Joy for All [JfA] cats and dogs). OBJECTIVE: We aimed to address the limited effectiveness research on JfA devices. METHODS: We conducted an 8-month, stratified, cluster randomized controlled trial in 8 care homes in Cornwall, United Kingdom. Over 4 months, 4 care homes each received 2 JfA devices (1 cat and 1 dog; intervention group), and 4 homes received care as usual (control group). Psychometrics were collected before and after the intervention to compare the change from baseline to follow-up between the groups. In the final 4 months, all 8 care homes received devices, but only qualitative data were collected owing to COVID-19 and reduced capacity. The primary outcome was neuropsychiatric symptoms (Neuropsychiatric Inventory [NPI] Nursing Home version). Care provider burden was a secondary outcome (occupational disruptiveness NPI subscale), alongside the Challenging Behavior scale, the Holden communication scale, the Campaign to End Loneliness questionnaire, and medication use. Qualitative data were collected through care staff observation calendars and end-of-study interviews to understand use, experience, and impact. We also collected demographic data and assessed dementia severity. In total, 253 residents had robot interaction opportunities, and 83 were consented for direct data collection. RESULTS: There was a significant difference in the total change from baseline to follow-up between the intervention and control groups for NPI (P<.001) and occupational disruptiveness (P=.03). Neuropsychiatric symptoms increased in the control group and decreased in the intervention group. No significant difference was seen for communication issues or challenging behavior. For NPI subdomains, there were significant differences from baseline to follow-up in delusions (P=.03), depression (P=.01), anxiety (P=.001), elation (P=.02), and apathy (P=.009), all of which decreased in the intervention group and increased slightly in the control group. The summative impact results suggested that most residents (46/54, 85%) who interacted with robots experienced a positive impact. Those who interacted had significantly higher dementia severity scores (P=.001). The qualitative results suggested good adoption, acceptability, and suitability for subjectively lonely individuals and lack of a novelty effect through sustained use, and demonstrated that the reasons for use were entertainment, anxiety, and agitation. CONCLUSIONS: Affordable robot pets hold potential for improving the well-being of care home residents and people with dementia, including reducing neuropsychiatric symptoms and occupational disruptiveness. This work suggests no novelty effect and contributes toward understanding robot pet suitability. Moreover, interactions were more common among residents with more moderate/severe dementia and those subjectively lonely. TRIAL REGISTRATION: ClinicalTrials.gov NCT04168463; https://www.clinicaltrials.gov/ct2/show/NCT04168463.

Voice-Enabled Intelligent Virtual Agents for People With Amnesia: Systematic Review.

BACKGROUND: Older adults often have increasing memory problems (amnesia), and approximately 50 million people worldwide have dementia. This syndrome gradually affects a patient over a period of 10-20 years. Intelligent virtual agents may support people with amnesia. OBJECTIVE: This study aims to identify state-of-the-art experimental studies with virtual agents on a screen capable of verbal dialogues with a target group of older adults with amnesia. METHODS: We conducted a systematic search of PubMed, SCOPUS, Microsoft Academic, Google Scholar, Web of Science, and CrossRef on virtual agent and amnesia on papers that describe such experiments. Search criteria were (Virtual Agent OR Virtual Assistant OR Virtual Human OR Conversational Agent OR Virtual Coach OR Chatbot) AND (Amnesia OR Dementia OR Alzheimer OR Mild Cognitive Impairment). Risk of bias was evaluated using the QualSyst tool (University of Alberta), which scores 14 study quality items. Eligible studies are reported in a table including country, study design type, target sample size, controls, study aims, experiment population, intervention details, results, and an image of the agent. RESULTS: A total of 8 studies was included in this meta-analysis. The average number of participants in the studies was 20 (SD 12). The verbal interactions were generally short. The usability was generally reported to be positive. The human utterance was seen in 7 (88%) out of 8 studies based on short words or phrases that were predefined in the agent's speech recognition algorithm. The average study quality score was 0.69 (SD 0.08) on a scale of 0 to 1. CONCLUSIONS: The number of experimental studies on talking about virtual agents that support people with memory problems is still small. The details on the verbal interaction are limited, which makes it difficult to assess the quality of the interaction and the possible effects of confounding parameters. In addition, the derivation of the aggregated data was difficult. Further research with extended and prolonged dialogues is required.

Kinematic-Based Classification of Social Gestures and Grasping by Humans and Machine Learning Techniques.

The affective motion of humans conveys messages that other humans perceive and understand without conventional linguistic processing. This ability to classify human movement into meaningful gestures or segments plays also a critical role in creating social interaction between humans and robots. In the research presented here, grasping and social gesture recognition by humans and four machine learning techniques (k-Nearest Neighbor, Locality-Sensitive Hashing Forest, Random Forest and Support Vector Machine) is assessed by using human classification data as a reference for evaluating the classification performance of machine learning techniques for thirty hand/arm gestures. The gestures are rated according to the extent of grasping motion on one task and the extent to which the same gestures are perceived as social according to another task. The results indicate that humans clearly rate differently according to the two different tasks. The machine learning techniques provide a similar classification of the actions according to grasping kinematics and social quality. Furthermore, there is a strong association between gesture kinematics and judgments of grasping and the social quality of the hand/arm gestures. Our results support previous research on intention-from-movement understanding that demonstrates the reliance on kinematic information for perceiving the social aspects and intentions in different grasping actions as well as communicative point-light actions.

User-Centered Design of Companion Robot Pets Involving Care Home Resident-Robot Interactions and Focus Groups With Residents, Staff, and Family: Qualitative Study.

BACKGROUND: Globally, pressure is increasing on health and social care resources due to the aging population and growing prevalence of dementia. Companion robots, such as Paro, demonstrate strong potential for helping reduce this pressure through reported benefits including reduced agitation, depression, loneliness, care provider burden, and medication use. However, we previously identified that user-centered design of robot pets is both essential and understudied. We observed that commonly used robot pets are poorly matched to end-user requirements, and that end users and developers of robot pets differ significantly in their perception of appropriate design. This may explain some of the contradictory outcome research and variance in results for robot pets, such as Paro. OBJECTIVE: In response to the literature gap, we aimed to provide user-centered insights into the design of robot pets from key stakeholders to inform future robot development and the choice of robots for real-world implementation and research. We focused on understanding user requirements. METHODS: We conducted a qualitative study with 65 participants from 5 care homes (26 care home residents, 29 staff members. and 10 family members). Care home residents formed groups of between 3 and 4 individuals and experienced free interactions with a range of 8 companion robots and toys, including Paro and more affordable alternatives. The robots provided had a range of esthetics, shell types, interactivity levels, and designs for comparison. Care staff and family members observed the interactions. All participants then engaged in focus groups within their stakeholder category to discuss preferences and user requirements in companion robot design. Both free interactions and focus groups were video and audio recorded, transcribed, and subjected to thematic analysis. RESULTS: Care home residents, family members, and staff were open and accepting of the use of companion robot pets, with the majority suggesting that they would keep a device for themselves or the residents. The most preferred device was the Joy for All cat, followed by the Joy for All dog. In discussions, the preferred design features included familiar animal embodiment (domestic pet), soft fur, interactivity, big appealing eyes, simulated breathing, and movements. Unfamiliar devices were more often seen as toy-like and suitable for children, producing some negative responses. CONCLUSIONS: This work provides important and user-centered insights into future robot designs for care home residents by means of a comprehensive comparison with key stakeholders. This work strongly supports the use of familiar embodiment in future robot pet designs, with domestic cat and dog morphologies appearing most acceptable. The results have implications for future robot designs and the selection of robot pets for both research and real-world implementations.

Emotional Dysregulation in Preschoolers with Autism Spectrum Disorder-A Sample of Romanian Children.

Emotional dysregulation problems seem to affect more than 80% of people with autism spectrum disorder (ASD) and may include irritability, aggressive behaviors, self-injury, and anxiety. Even though these types of problems are very common and affect the well-being of individuals with ASD, there are no objective assessment tools developed for this population and there are only a few intervention techniques meant to address these symptoms. This study investigates the feasibility of using off-the-shelf wearable devices to accurately measure heart rate, which has been associated with emotional dysregulation, and to test the effectiveness of functional communication training in reducing the emotional outburst in preschoolers with ASD. We used a single-case experiment design with three preschoolers with ASD to test if the duration of the emotional outburst and the elevated heart rate levels can be reduced by using functional communication training. Our results show that for two of the participants, the intervention was effective in reducing the duration of behaviors associated with emotional outburst, and that there were significant differences between baseline and intervention phase in terms of heart rate levels. However, our results are inconclusive regarding the association between elevated heart rates and the occurrence of the emotional outburst. Nevertheless, more research is needed to investigate the use of off-the-shelf wearable devices in predicting challenging behaviors in children with ASD.

The DREAM Dataset: Supporting a data-driven study of autism spectrum disorder and robot enhanced therapy.

We present a dataset of behavioral data recorded from 61 children diagnosed with Autism Spectrum Disorder (ASD). The data was collected during a large-scale evaluation of Robot Enhanced Therapy (RET). The dataset covers over 3000 therapy sessions and more than 300 hours of therapy. Half of the children interacted with the social robot NAO supervised by a therapist. The other half, constituting a control group, interacted directly with a therapist. Both groups followed the Applied Behavior Analysis (ABA) protocol. Each session was recorded with three RGB cameras and two RGBD (Kinect) cameras, providing detailed information of children's behavior during therapy. This public release of the dataset comprises body motion, head position and orientation, and eye gaze variables, all specified as 3D data in a joint frame of reference. In addition, metadata including participant age, gender, and autism diagnosis (ADOS) variables are included. We release this data with the hope of supporting further data-driven studies towards improved therapy methods as well as a better understanding of ASD in general.

Microbial contamination and efficacy of disinfection procedures of companion robots in care homes.

BACKGROUND: Paro and other robot animals can improve wellbeing for older adults and people with dementia, through reducing depression, agitation and medication use. However, nursing and care staff we contacted expressed infection control concerns. Little related research has been published. We assessed (i) how microbiologically contaminated robot animals become during use by older people within a care home and (ii) efficacy of a cleaning procedure. METHODS: This study had two stages. In stage one we assessed microbial load on eight robot animals after interaction with four care home residents, and again following cleaning by a researcher. Robot animals provided a range of shell-types, including fur, soft plastic, and solid plastic. Stage two involved a similar process with two robot animals, but a care staff member conducted cleaning. The cleaning process involved spraying with anti-bacterial product, brushing fur-type shells, followed by vigorous top-to-tail cleaning with anti-bacterial wipes on all shell types. Two samples were taken from each of eight robots in stage one and two robots in stage two (20 samples total). Samples were collected using contact plate stamping and evaluated using aerobic colony count and identification (gram stain, colony morphology, coagulase agglutination). Colony counts were measured by colony forming units per square centimetre (CFU/cm2). RESULTS: Most robots acquired microbial loads well above an acceptable threshold of 2.5 CFU/cm2 following use. The bacteria identified were micrococcus species, coagulase negative staphylococcus, diptheriods, aerobic spore bearers, and staphylococcus aureus, all of which carry risk for human health. For all devices the CFU/cm2 reduced to well within accepted limits following cleaning by both researcher and care staff member. CONCLUSIONS: Companion robots will acquire significant levels of bacteria during normal use. The simple cleaning procedure detailed in this study reduced microbial load to acceptable levels in controlled experiments. Further work is needed in the field and to check the impact on the transmission of viruses.

Ethical perceptions towards real-world use of companion robots with older people and people with dementia: survey opinions among younger adults.

BACKGROUND: Use of companion robots may reduce older people's depression, loneliness and agitation. This benefit has to be contrasted against possible ethical concerns raised by philosophers in the field around issues such as deceit, infantilisation, reduced human contact and accountability. Research directly assessing prevalence of such concerns among relevant stakeholders, however, remains limited, even though their views clearly have relevance in the debate. For example, any discrepancies between ethicists and stakeholders might in itself be a relevant ethical consideration while concerns perceived by stakeholders might identify immediate barriers to successful implementation. METHODS: We surveyed 67 younger adults after they had live interactions with companion robot pets while attending an exhibition on intimacy, including the context of intimacy for older people. We asked about their perceptions of ethical issues. Participants generally had older family members, some with dementia. RESULTS: Most participants (40/67, 60%) reported having no ethical concerns towards companion robot use when surveyed with an open question. Twenty (30%) had some concern, the most common being reduced human contact (10%), followed by deception (6%). However, when choosing from a list, the issue perceived as most concerning was equality of access to devices based on socioeconomic factors (m = 4.72 on a scale 1-7), exceeding more commonly hypothesized issues such as infantilising (m = 3.45), and deception (m = 3.44). The lowest-scoring issues were potential for injury or harm (m = 2.38) and privacy concerns (m = 2.17). Over half (39/67 (58%)) would have bought a device for an older relative. Cost was a common reason for choosing not to purchase a device. CONCLUSIONS: Although a relatively small study, we demonstrated discrepancies between ethical concerns raised in the philosophical literature and those likely to make the decision to buy a companion robot. Such discrepancies, between philosophers and 'end-users' in care of older people, and in methods of ascertainment, are worthy of further empirical research and discussion. Our participants were more concerned about economic issues and equality of access, an important consideration for those involved with care of older people. On the other hand the concerns proposed by ethicists seem unlikely to be a barrier to use of companion robots.

Probing the early phase of rapid instructed rule encoding.

Humans can rapidly convert instructions about a rule into functional neural structures used to apply the rule. The early stages of this encoding process are poorly understood. We designed a stimulus-response (SR) task in which participants were first shown a SR rule on a screen for 200 ms, and then had to apply it to a test stimulus T, which either matched the S in the rule (SR trial) or not (catch trial). To investigate the early stages of rule encoding, the delay between the end of rule display and the onset of the test stimulus was manipulated and chosen between values of 50 ms to 1300 ms. Participants conducted three sessions of 288 trials each, separated by a median of 9 h. Random sequences of 20 rules were used. We then analysed the reaction times and the types of errors made by participants in the different conditions. The analysis of practice effects in session 1 suggests that the neural networks that process SR and catch trials are at least partially distinct, and improve separately during the practice of respectively SR and catch trials. The rule-encoding process, however, is common to both tasks and improves with the number of trials, irrespective of the trial type. Rule encoding shows interesting dynamic properties that last for 500 ms after the end of the stimulus presentation. The encoding process increases the response time in a non-stochastic way, simply adding a reaction time cost to all responses. The rule-retrieval system is functional before the encoding has stabilized, as early as 50 ms after the end of SR rule presentation, with low response errors. It is sensitive to masking however, producing errors with brief (100 ms) test stimulus presentations. Once encoding has stabilized, the sensitivity to masking disappears. It is suggested that participants do encode rules as a parametrized function, using the same neural encoding structure for each trial, rather than reconfiguring their brain anew for each new SR rule. This structure would have been implemented from instructions received prior to the experiment, by using a library of neural functions available in the brain. The observed errors are consistent with this view.

Companion robots for older people: importance of user-centred design demonstrated through observations and focus groups comparing preferences of older people and roboticists in South West England.

OBJECTIVE: Companion robots, such as Paro, may reduce agitation and depression for older people with dementia. However, contradictory research outcomes suggest robot design is not always optimal. While many researchers suggest user-centred design is important, there is little evidence on the difference this might make. Here, we aimed to assess its importance by comparing companion robot design perceptions between older people (end users) and roboticists (developers). DESIGN: Older people and roboticists interacted with eight companion robots or alternatives at two separate events in groups of two to four people. Interactions were recorded, participants' comments and observations were transcribed, and content was analysed. Subsequently, each group participated in focus groups on perceptions of companion robot design. Discussions were recorded and transcribed, and content was analysed. PARTICIPANTS AND SETTINGS: Seventeen older people (5 male, 12 female, ages 60-99) at a supported living retirement complex, and 18 roboticists (10 male, 8 female, ages 24-37) at a research centre away-day. RESULTS: We found significant differences in design preferences between older people and roboticists. Older people desired soft, furry, interactive animals that were familiar and realistic, while unfamiliar forms were perceived as infantilising. By contrast, most roboticists eschewed familiar and realistic designs, thinking unfamiliar forms better suited older people. Older people also expressed desire for features not seen as important by developers. A large difference was seen in attitude towards ability to talk: 12/17 (71%) older people but only 2/18 (11%) roboticists requested speech. Older people responded positively towards life-simulation features, eye contact, robot personalisation and obeying commands, features undervalued by roboticists. These differences were reflected in preferred device, with 'Joy for All' cat chosen most often by older people, while roboticists most often chose Paro. CONCLUSION: The observed misalignment of opinion between end users and developers on desirable design features of companion robots demonstrates the need for user-centred design during development.

What Can You See? Identifying Cues on Internal States From the Movements of Natural Social Interactions.

In recent years, the field of Human-Robot Interaction (HRI) has seen an increasing demand for technologies that can recognize and adapt to human behaviors and internal states (e.g., emotions and intentions). Psychological research suggests that human movements are important for inferring internal states. There is, however, a need to better understand what kind of information can be extracted from movement data, particularly in unconstrained, natural interactions. The present study examines which internal states and social constructs humans identify from movement in naturalistic social interactions. Participants either viewed clips of the full scene or processed versions of it displaying 2D positional data. Then, they were asked to fill out questionnaires assessing their social perception of the viewed material. We analyzed whether the full scene clips were more informative than the 2D positional data clips. First, we calculated the inter-rater agreement between participants in both conditions. Then, we employed machine learning classifiers to predict the internal states of the individuals in the videos based on the ratings obtained. Although we found a higher inter-rater agreement for full scenes compared to positional data, the level of agreement in the latter case was still above chance, thus demonstrating that the internal states and social constructs under study were identifiable in both conditions. A factor analysis run on participants' responses showed that participants identified the constructs interaction imbalance, interaction valence and engagement regardless of video condition. The machine learning classifiers achieved a similar performance in both conditions, again supporting the idea that movement alone carries relevant information. Overall, our results suggest it is reasonable to expect a machine learning algorithm, and consequently a robot, to successfully decode and classify a range of internal states and social constructs using low-dimensional data (such as the movements and poses of observed individuals) as input.

Representational fluidity in embodied (artificial) cognition.

Theories of embodied cognition agree that the body plays some role in human cognition, but disagree on the precise nature of this role. While it is (together with the environment) fundamentally engrained in the so-called 4E (or multi-E) cognition stance, there also exists interpretations wherein the body is merely an input/output interface for cognitive processes that are entirely computational. In the present paper, we show that even if one takes such a strong computationalist position, the role of the body must be more than an interface to the world. To achieve human cognition, the computational mechanisms of a cognitive agent must be capable not only of appropriate reasoning over a given set of symbolic representations; they must in addition be capable of updating the representational framework itself (leading to the titular representational fluidity). We demonstrate this by considering the necessary properties that an artificial agent with these abilities need to possess. The core of the argument is that these updates must be falsifiable in the Popperian sense while simultaneously directing representational shifts in a direction that benefits the agent. We show that this is achieved by the progressive, bottom-up symbolic abstraction of low-level sensorimotor connections followed by top-down instantiation of testable perception-action hypotheses. We then discuss the fundamental limits of this representational updating capacity, concluding that only fully embodied learners exhibiting such a priori perception-action linkages are able to sufficiently ground spontaneously-generated symbolic representations and exhibit the full range of human cognitive capabilities. The present paper therefore has consequences both for the theoretical understanding of human cognition, and for the design of autonomous artificial agents.

Affordance processing in segregated parieto-frontal dorsal stream sub-pathways.

The concept of affordances indicates "action possibilities" as characterized by object properties the environment provides to interacting organisms. Affordances relate to both perception and action and refer to sensory-motor processes emerging from goal-directed object interaction. In contrast to stable properties, affordances may vary with environmental context. A sub-classification into stable and variable affordances was proposed in the framework of the ROSSI project (Borghi et al., 2010; Borghi and Riggio, 2015, 2009). Here, we present a coordinate-based meta-analysis of functional imaging studies on object interaction targeting consistent anatomical correlates of these different types of affordances. Our review revealed the existence of two parallel (but to some extent overlapping) functional pathways. The network for stable affordances consists of predominantly left inferior parietal and frontal cortices in the ventro-dorsal stream, whereas the network for variable affordances is localized preferentially in the dorso-dorsal stream. This is in line with the proposal of differentiated affordances: stable affordances are characterized by the knowledge of invariant object features, whereas variable affordances underlie adaptation to changing object properties.

What's on the Inside Counts: A Grounded Account of Concept Acquisition and Development.

Understanding the factors which affect the age of acquisition (AoA) of words and concepts is fundamental to understanding cognitive development more broadly. Traditionally, studies of AoA have taken two approaches, either exploring the effect of linguistic variables such as input frequency (e.g., Naigles and Hoff-Ginsberg, 1998) or the semantics of the underlying concept, such as concreteness or imageability (e.g., Bird et al., 2001). Embodied theories of cognition, meanwhile, assume that concepts, even relatively abstract ones, can be grounded in the embodied experience. While the focus of such discussions has been mainly on grounding in external modalities, more recently some have argued for the importance of interoceptive features, or grounding in complex modalities such as social interaction. In this paper, we argue for the integration and extension of these two strands of research. We demonstrate that the psycholinguistic factors traditionally considered to determine AoA are far from sufficient to account for the variability observed in AoA data. Given this gap, we propose groundability as a new conceptual tool that can measure the degree to which concepts are grounded both in external and, critically, internal modalities. We then present a mechanistic theory of conceptual representation that can account for groundability in addition to the existing variables argued to influence concept acquisition in both the developmental and embodied cognition literatures, and discuss its implications for future work in concept and cognitive development.

Embodied cognition and circular causality: on the role of constitutive autonomy in the reciprocal coupling of perception and action.

The reciprocal coupling of perception and action in cognitive agents has been firmly established: perceptions guide action but so too do actions influence what is perceived. While much has been said on the implications of this for the agent's external behavior, less attention has been paid to what it means for the internal bodily mechanisms which underpin cognitive behavior. In this article, we wish to redress this by reasserting that the relationship between cognition, perception, and action involves a constitutive element as well as a behavioral element, emphasizing that the reciprocal link between perception and action in cognition merits a renewed focus on the system dynamics inherent in constitutive biological autonomy. Our argument centers on the idea that cognition, perception, and action are all dependent on processes focussed primarily on the maintenance of the agent's autonomy. These processes have an inherently circular nature-self-organizing, self-producing, and self-maintaining-and our goal is to explore these processes and suggest how they can explain the reciprocity of perception and action. Specifically, we argue that the reciprocal coupling is founded primarily on their endogenous roles in the constitutive autonomy of the agent and an associated circular causality of global and local processes of self-regulation, rather than being a mutual sensory-motor contingency that derives from exogenous behavior. Furthermore, the coupling occurs first and foremost via the internal milieu realized by the agent's organismic embodiment. Finally, we consider how homeostasis and the related concept of allostasis contribute to this circular self-regulation.

On the importance of a rich embodiment in the grounding of concepts: perspectives from embodied cognitive science and computational linguistics.

The recent trend in cognitive robotics experiments on language learning, symbol grounding, and related issues necessarily entails a reduction of sensorimotor aspects from those provided by a human body to those that can be realized in machines, limiting robotic models of symbol grounding in this respect. Here, we argue that there is a need for modeling work in this domain to explicitly take into account the richer human embodiment even for concrete concepts that prima facie relate merely to simple actions, and illustrate this using distributional methods from computational linguistics which allow us to investigate grounding of concepts based on their actual usage. We also argue that these techniques have applications in theories and models of grounding, particularly in machine implementations thereof. Similarly, considering the grounding of concepts in human terms may be of benefit to future work in computational linguistics, in particular in going beyond "grounding" concepts in the textual modality alone. Overall, we highlight the overall potential for a mutually beneficial relationship between the two fields.

Theories and computational models of affordance and mirror systems: an integrative review.

Neuroscientific and psychological data suggest a close link between affordance and mirror systems in the brain. However, we still lack a full understanding of both the individual systems and their interactions. Here, we propose that the architecture and functioning of the two systems is best understood in terms of two challenges faced by complex organisms, namely: (a) the need to select among multiple affordances and possible actions dependent on context and high-level goals and (b) the exploitation of the advantages deriving from a hierarchical organisation of behaviour based on actions and action-goals. We first review and analyse the psychological and neuroscientific literature on the mechanisms and processes organisms use to deal with these challenges. We then analyse existing computational models thereof. Finally we present the design of a computational framework that integrates the reviewed knowledge. The framework can be used both as a theoretical guidance to interpret empirical data and design new experiments, and to design computational models addressing specific problems debated in the literature.