This paper proposes an approach to investigate norm-governed learning agents which combines a logic-based formalism with an equation-based counterpart. This dual formalism enables us to describe the reasoning of such agents and their interactions using argumentation, and, at the same time, to capture systemic features using equations. The approach is applied to norm emergence and internalisation in systems of learning agents. The logical formalism is rooted into a probabilistic defeasible logic instantiating Dung’s argumentation framework. Rules of this logic are attached (...) with probabilities to describe the agents’ minds and behaviours as well as uncertain environments. Then, the equation-based model for reinforcement learning, defined over this probability distribution, allows agents to adapt to their environment and self-organise. (shrink)

In the past decade, the fields of machine learning and artificial intelligence (AI) have seen unprecedented developments that raise human-machine interactions (HMI) to the next level.Smart machines, i.e., machines endowed with artificially intelligent systems, have lost their character as mere instruments. This, at least, seems to be the case if one considers how humans experience their interactions with them. Smart machines are construed to serve complex functions involving increasing degrees of freedom, and they generate solutions not fully anticipated by humans. (...) Consequently, their performances show a touch of action and even autonomy. HMI is therefore often described as a sort of “cooperation” rather than as a mere application of a tool. Some authors even go as far as subsuming cooperation with smart machines under the label ofpartnership, akin to cooperation between human agents sharing a common goal. In this paper, we explore how far the notion of shared agency and partnership can take us in our understanding of human interaction with smart machines. Discussing different topoi related to partnerships in general, we suggest that different kinds of “partnership” depending on the form of interaction between agents need to be kept apart. Building upon these discussions, we propose a tentative taxonomy of different kinds of HMI distinguishing coordination, collaboration, cooperation, and social partnership. (shrink)

In this contribution we analyze how the actions of rational agents feed back on their beliefs. We present two agent-based computer simulations studying complex social interactions in which agents that follow utility maximizing strategies thereby deteriorate their own long-term quality of beliefs. We take these results as a starting point to discuss the complex relationship between rational action couched in terms of maximizing utility and the emergence of informational inequalities.

Interactions between an intelligent software agent and a human user are ubiquitous in everyday situations such as access to information, entertainment, and purchases. In such interactions, the ISA mediates the user’s access to the content, or controls some other aspect of the user experience, and is not designed to be neutral about outcomes of user choices. Like human users, ISAs are driven by goals, make autonomous decisions, and can learn from experience. Using ideas from bounded rationality, we frame these (...) interactions as instances of an ISA whose reward depends on actions performed by the user. Such agents benefit by steering the user’s behaviour towards outcomes that maximise the ISA’s utility, which may or may not be aligned with that of the user. Video games, news recommendation aggregation engines, and fitness trackers can all be instances of this general case. Our analysis facilitates distinguishing various subcases of interaction, as well as second-order effects that might include the possibility for adaptive interfaces to induce behavioural addiction, and/or change in user belief. We present these types of interaction within a conceptual framework, and review current examples of persuasive technologies and the issues that arise from their use. We argue that the nature of the feedback commonly used by learning agents to update their models and subsequent decisions could steer the behaviour of human users away from what benefits them, and in a direction that can undermine autonomy and cause further disparity between actions and goals as exemplified by addictive and compulsive behaviour. We discuss some of the ethical, social and legal implications of this technology and argue that it can sometimes exploit and reinforce weaknesses in human beings. (shrink)

This study investigates the acquisition of integrated object manipulation and categorization abilities through a series of experiments in which human adults and artificial agents were asked to learn to manipulate two-dimensional objects that varied in shape, color, weight, and color intensity. The analysis of the obtained results and the comparison of the behavior displayed by human and artificial agents allowed us to identify the key role played by features affecting the agent/environment interaction, the relation between category and action development, (...) and the role of cognitive biases originating from previous knowledge. (shrink)

The Internet of Things (IoT) is a fast-growing technology that connects everyday devices to the Internet, enabling wireless, low-consumption and low-cost communication and data exchange. IoT has revolutionized the way devices interact with each other and the internet. The more devices become connected, the greater the risk of security breaches. There is currently a need for new approaches to algorithms that can detect malware regardless of the size of the network and that can adapt to dynamic changes in the network. (...) Through the use of a multi-agent reinforcement learning algorithm, this paper proposes a novel algorithm for malware detection in IoT devices. The proposed algorithm is not strongly dependent on the size of the IoT network due to the that its training is adapted using time differences if the IoT network size is small or Monte Carlo otherwise. To validate the proposed algorithm in an environment as close to reality as possible, we proposed a scenario based on a real IoT network, where we tested different malware propagation models. Different simulations varying the number of agents and nodes in the IoT network have been developed. The result of these simulations proves the efficiency and adaptability of the proposed algorithm in detecting malware, regardless of the malware propagation model. (shrink)

This paper considers a new class of agent dynamic logics which provide a formal means of specifying and reasoning about the agents activities and informational, motivational and practical aspects of the behaviour of the agents. We present a Hilbert-style deductive system for a basic agent dynamic logic and consider a number of extensions of this logic with axiom schemata formalising interactions between knowledge and commitment (expressing an agent s awareness of her commitments), and interactions between knowledge and (...) actions (expressing no learning and persistence of knowledge after actions). The deductive systems are proved sound and complete with respect to a Kripke-style semantics. Each of the considered logics is shown to have the small model property and therefore decidable. (shrink)

Diversity of agents occurs naturally in epistemic logic, and dynamic logics of information update and belief revision. In this paper we provide a systematic discussion of different sources of diversity, such as introspection ability, powers of observation, memory capacity, and revision policies, and we show how these can be encoded in dynamic epistemic logics allowing for individual variation among agents. Next, we explore the interaction of diverse agents by looking at some concrete scenarios of communication and learning, and we propose (...) a logical methodology to deal with these as well. We conclude with some further questions on the logic of diversity and interaction. (shrink)

Temporal epistemic logics are known, from results of Halpern and Vardi, to have a wide range of complexities of the satisfiability problem: from PSPACE, through non-elementary, to highly undecidable. These complexities depend on the choice of some key parameters specifying, inter alia, possible interactions between time and knowledge, such as synchrony and agents' abilities for learning and recall. In this work we develop practically implementable tableau-based decision procedures for deciding satisfiability in single-agent synchronous temporal-epistemic logics with interactions between time (...) and knowledge. We discuss some complications that occur, even in the single-agent case, when interactions between time and knowledge are assumed and show how the method of incremental tableaux can be adapted to work in EXPSPACE, respectively 2EXPTIME, for these logics, thereby also matching the upper bounds obtained for them by Halpern and Vardi. (shrink)

The results of a series of computer simulations demonstrate how the introduction of separate spatial dimensions for agent interaction and learning respectively affects the possibility of cooperation evolving in the repeated prisoner's dilemma played by populations of boundedly-rational agents. In particular, the localisation of learning promotes the emergence of cooperative behaviour, while the localisation of interaction has an ambiguous effect on it.

The study of imitation and other mechanisms of social learning is an exciting area of research for all those interested in understanding the origin and the nature of animal learning in asocial context. Moreover, imitation is an increasingly important research topic in Artificial Intelligence and social robotics which opens up the possibility ofindividualized social intelligencein robots that are part of a community, and allows us to harness not only individual learning by the single robot, but also the acquisition of new (...) skills by observing other members of the community. After an introduction to the main research issues in research on imitation in various fields, we motivate the particular focus of this work, namely thecorrespondence problem. We describeAction Learning for Imitation via Correspondences between Embodiments, an implemented generic framework for solving the correspondence problem between differently embodied robots. Alice enables a robotic agent to learn a behavioral repertoire suitable to performing a task by observing a model agent. Importantly, the model agent could possibly possess a different type of body, e.g. a different number of limbs or joints, a different height, different sensors, a different basic action repertoire, etc. Previously, in a test-bed where the agents differed according to their possible movement patterns, we demonstrated that the character of imitation achieved will depend on the granularity of subgoal matching, and on the metrics used to evaluate success.In our current work, we implemented Alice in a new test-bed calledRabitwhere simple simulated robotic arm agents use various metrics for evaluating success according to actions, states, effects, or weighted combinations.We examine the roles ofsynchronization,looseness of perceptual matching, and ofproprioceptive matchingby a series of experiments. Also, we study how Alice copes withchanges in the embodimentof the imitator during learning. Our simulation results suggest thatsynchronizationandloose perceptual matchingallow for faster acquisition of behavioral compentencies at low error rates.Social learning plays a role as areplicationmechanism for behaviors and results invariabilitywhen the transmitted behavior differs from the model’s behavior, thus providing theevolutionary substrate for cultureand its pre-cursors. Social learning in robotics could therefore serve as the basis for culture in societies whose members include artificial agents. We address the use of imitative social learning mechanisms like Alice for transmission of skills between robots, and give first examples of transmission of a skill despite differences in embodiment of agents involved. In the particular setup, transmission occurs through a chain, as well as emerging in cyclic arrangements of robots. These simple examples demonstrate that by using social learning and imitation,cultural transmissionis possible among robots, even in heterogeneous groups of robots. (shrink)

Co-learning of multiple agents has been studied in co-learning settings, and how do they help, or many different disciplines under various guises. For hamper, learning and cooperation? example, the issue has been tackled by distributed • How do we characterize the process and the artificial intelligence, parallel and distributed com- dynamics of co-learning, conceptually, mathe- puting, cognitive psychology, social psychology, matically, or computationally? game theory (and other areas of mathematical econ- • how do social structures and relations interact omics), sociology, (...) anthropology, and many other with co-learning of multiple agents? related disciplines. (shrink)

We present an approach to subjective computing for the design of future robots that exhibit more adaptive and flexible behavior in terms of subjective intelligence. Instead of encapsulating subjectivity into higher order states, we show by means of a relational approach how subjective intelligence can be implemented in terms of the reciprocity of autonomous self-referentiality and direct world-coupling. Subjectivity concerns the relational arrangement of an agent’s cognitive space. This theoretical concept is narrowed down to the problem of coaching a (...) reinforcement learning agent by means of binary feedback. Algorithms are presented that implement subjective computing. The relational characteristic of subjectivity is further confirmed by a questionnaire on human perception of robot’s behavior. The results imply that subjective intelligence cannot be externally observed. In sum, we conclude that subjective intelligence in relational terms is fully tractable and therefore implementable in artificial agents. (shrink)

Context: The advent of a general artificial intelligence mechanism that learns like humans do would represent the realization of an old and major dream of science. It could be achieved by an artifact able to develop its own cognitive structures following constructivist principles. However, there is a large distance between the descriptions of the intelligence made by constructivist theories and the mechanisms that currently exist. Problem: The constructivist conception of intelligence is very powerful for explaining how cognitive development takes place. (...) However, until now, no computational model has successfully demonstrated the underlying mechanisms necessary to realize it. In other words, the artificial intelligence (AI) community has not been able to give rise to a system that convincingly implements the principles of intelligence as postulated by constructivism, and that is also capable of dealing with complex environments. Results: This paper presents the constructivist anticipatory learning mechanism (CALM), an agent learning mechanism based on the constructivist approach of AI. It is designed to deal dynamically and interactively with environments that are at the same time partially deterministic and partially observable. CALM can model the regularities experienced in the interaction with the environment, on the sensorimotor level as well, as by constructing abstract or high-level representational concepts. The created model provides the knowledge necessary to generate the agent behavior. The paper also presents the coupled agent environment system (CAES) meta-architecture, which defines a conception of an autonomous agent, situated in the environment, embodied and intrinsically motivated. Implications: The paper can be seen as a step towards a computational implementation of constructivist principles, on the one hand suggesting a further perspective of this refreshing movement on the AI field (which is still too steeped in a behaviorist influence and dominated by probabilistic models and narrow applied approaches), and on the other hand bringing some abstract descriptions of the cognitive process into a more concrete dimension, in the form of algorithms. Constructivist content: The connection of this paper with constructivism is the proposal of a computational and formally described mechanism that implements important aspects of the subjective process of knowledge construction based on key ideas proposed by constructivist theories. (shrink)

This article introduces a computational model which illustrates how a population of agents can coordinate a vocabulary for goal oriented behavior through repeated local interactions, called “Action Games”. Using principles of self organization and specific assumptions on their behavior, the agents learn the goals and a vocabulary for them. It is shown that the proposed model can be used to investigate the coordination of vocabularies for goal oriented behavior both in a vertical and in a horizontal transmission scheme. Furthermore, it (...) is shown that in this model shared vocabularies can only be learned from multiple demonstrations consisting of exactly these actions that are strictly required to reach the goal. Keywords: robot imitation; self organization; vocabularies. (shrink)

This article introduces a computational model which illustrates how a population of agents can coordinate a vocabulary for goal oriented behavior through repeated local interactions, called “Action Games”. Using principles of self organization and specific assumptions on their behavior, the agents learn the goals and a vocabulary for them. It is shown that the proposed model can be used to investigate the coordination of vocabularies for goal oriented behavior both in a vertical and in a horizontal transmission scheme. Furthermore, it (...) is shown that in this model shared vocabularies can only be learned from multiple demonstrations consisting of exactly these actions that are strictly required to reach the goal. Keywords: robot imitation; self organization; vocabularies. (shrink)

The role of interaction in learning is essential and profound: it must provide the means to solve open problems (those only vaguely specified in advance), but cannot be captured using our familiar formal cognitive tools. This presents an impasse to those confined to present formalisms; but interaction is fundamentally dynamical, not formal, and with its importance thus underlined it invites the development of a distinctively interactivist account of life and mind. This account is provided, from its roots in the interactivist (...) biological constitution of life, through the evolution of the dual internal regulatory capacities expressed as intentionality and intelligence, to its expression in self-directed anticipative learning in persons and in science. (shrink)

Economies are complicated systems encompassing micro behaviors, interaction patterns, and global regularities. Whether partial or general in scope, studies of economic systems must consider how to handle difficult real-world aspects such as asymmetric information, imperfect competition, strategic interaction, collective learning, and the possibility of multiple equilibria. Recent advances in analytical and computational tools are permitting new approaches to the quantitative study of these aspects. One such approach is Agent-based Computational Economics (ACE), the computational study of economic processes modeled as (...) dynamic systems of interacting agents. This chapter explores the potential advantages and disadvantages of ACE for the study of economic systems. General points are concretely illustrated using an ACE model of a two-sector decentralized market economy. Six issues are highlighted: Constructive understanding of production, pricing, and trade processes; the essential primacy of survival; strategic rivalry and market power; behavioral uncertainty and learning; the role of conventions and organizations; and the complex interactions among structural attributes, institutional arrangements, and behavioral dispositions. (shrink)

Recent studies in pedestrian simulation have been able to construct a highly realistic navigation behaviour in many circumstances. However, when replicating the close interactions between pedestrians, the replicated behaviour is often unnatural and lacks human likeness. One of the possible reasons is that the current models often ignore the cognitive factors in the human thinking process. Another reason is that many models try to approach the problem by optimising certain objectives. On the other hand, in real life, humans do not (...) always take the most optimised decisions, particularly when interacting with other people. To improve the navigation behaviour in this circumstance, we proposed a pedestrian interacting model using reinforcement learning. Additionally, a novel cognitive prediction model, inspired by the predictive system of human cognition, is also incorporated. This helps the pedestrian agent in our model to learn to interact and predict the movement in a similar practice as humans. In our experimental results, when compared to other models, the path taken by our model’s agent is not the most optimised in certain aspects like path lengths, time taken and collisions. However, our model is able to demonstrate a more natural and human-like navigation behaviour, particularly in complex interaction settings. (shrink)

The energy domain is still dominated by equilibrium models that underestimate both the dangers and opportunities related to climate change. In reality, climate and energy systems contain tipping points, feedback loops, and exponential developments. This paper describes how to create realistic energy transition management models: quantitative models that can discover profitable pathways from fossil fuels to renewable energy. We review the literature regarding agent-based economics, disruptive innovation, and transition management and determine the following requirements. Actors must be detailed, heterogeneous, (...) interacting, learning, and strategizing. Technology should be represented as a detailed and heterogeneous portfolio that can develop in a bottom-up manner, using endogenous feedback loops. Assumptions about discount rates and the social cost of carbon should be configurable. The model should contain interactions between the global, national, local, and individual level. A review of modelling techniques shows that equilibrium models are unsuitable and that system dynamics and discrete event simulation are too limited. The agent-based approach is found to be uniquely suited for the complex adaptive sociotechnical systems that must be modelled. But the choice for agent-based models does not mean a rejection of other approaches because they can be accommodated within the agent-based framework. We conclude with practical guidelines. (shrink)

Machine learning researchers distinguish between reinforcement learning and supervised learning and refer to reinforcement learning systems as “agents”. This paper vindicates the claim that systems trained by reinforcement learning are agents while those trained by supervised learning are not. Systems of both kinds satisfy Dretske’s criteria for agency, because they both learn to produce outputs selectively in response to inputs. However, reinforcement learning is sensitive to the instrumental value of outputs, giving rise to systems which exploit the effects of outputs (...) on subsequent inputs to achieve good performance over episodes of interaction with their environments. Supervised learning systems, in contrast, merely learn to produce better outputs in response to individual inputs. (shrink)

This article investigates agency as a way of being and acting in sustainability leadership. Our primary aim is to enhance understanding of agentic strategies that facilitate transcending systemic complexities in sustainability leadership. We make a distinction in our analytical approach by drawing from Emirbayer and Mische’s conceptualisation of agency as ‘an interactive process of reflexive transformation and relational pragmatics, a temporally embedded process of social engagement, informed by the past, oriented towards the future and enacted in the present’ . We (...) add ontological sources of agency to these dynamics which interact with habit, imagination, judgement and learning in the transformation of social systems. This approach underpins our model ‘Dynamic Interactions of Agency in Leadership’ , an integrative framework for analysing agency in sustainability leadership. We examine the efficacy of our framework in higher education initiatives in which sustainability aspirations, aims and actions are envisioned, articulated and mobilised. We conclude by offering further avenues of research in sustainability leadership designed to advance this burgeoning field and contribute to bridging the gap between sustainability challenges and our abilities to solve them. (shrink)

In species capable of learning, including our own, individuals can modify their behavior by some adaptive process. Important classes of behavior - mating, predation, coalitions, trade, signaling, and division of labor - involve interactions between individuals. The agents involved learn two things: with whom to interact and how to act. That is to say that adaptive dynamics operates both on structure and strategy.

Formal epistemology is the study of crucial concepts in general or main- stream epistemology including knowledge, belief , certainty, ra- tionality, reasoning, decision, justi cation, learning, agent interaction and information processing using a spread of di¤erent formal tools. These formal tools may be drawn from elds such as logic, probability theory, game theory, decision theory, formal learning theory, and distributed com- puting –such variety is typical in formal epistemology, a eld in which interaction with topics outside of philosophy proper (...) is the rule rather than the exception. Practitioners of formal epistemology include philosophers, computer scientists, social scientists, cognitive psychologists, theoretical economists, mathematicians, and theoretical linguists. The interdiscipli- nary nature of formal epistemology can make it di¢ cult for those new to the eld to have a sense of some of its basic agendas, actors, and issues. What follows is a breezy overview of formal epistemology as organized around notions of agency and interaction. (shrink)

Open peer commentary on the article “Subsystem Formation Driven by Double Contingency” by Bernd Porr & Paolo Di Prodi. Upshot: Agents that learn from perturbations of closed control loops are considered constructivist by virtue of the fact that their input (the perturbation) does not convey ontological information about the environment. That is, they learn by actively experiencing their environment through interaction, as opposed to learning by registering directly input data characterizing the environment. Generalizing this idea, the notion of learning by (...) experiencing provides a broader conceptual framework than cybernetic control theory for studying the double contingency problem, and may yield more progress in constructivist agent design. (shrink)

Young children generally learn words from other people. Recent research has shown that children can learn new actions and skills from nonhuman agents. This study examines whether young children could learn words from a robot. Preschool children were shown a video in which either a woman or a mechanical robot labeled novel objects. Then the children were asked to select the objects according to the names used in the video. The results revealed that children in the human condition were more (...) likely to select the correct objects than those in the robot condition. Nevertheless, the five-year-old children in the robot condition performed significantly better than chance level, while the four-year olds did not. Thus there is a developmental difference in children’s potential to learn words from a robot. The results contribute to our understanding of how children interact with non-human agents. Keywords: developmental cybernetics; word learning; social cognition; cognitive development. (shrink)

The area of smart power grids needs to constantly improve its efficiency and resilience, to provide high quality electrical power in a resilient grid, while managing faults and avoiding failures. Achieving this requires high component reliability, adequate maintenance, and a studied failure occurrence. Correct system operation involves those activities and novel methodologies to detect, classify, and isolate faults and failures and model and simulate processes with predictive algorithms and analytics. In this paper, we showcase the application of a complex-adaptive, self-organizing (...) modeling method, and Probabilistic Boolean Networks, as a way towards the understanding of the dynamics of smart grid devices, and to model and characterize their behavior. This work demonstrates that PBNs are equivalent to the standard Reinforcement Learning Cycle, in which the agent/model has an interaction with its environment and receives feedback from it in the form of a reward signal. Different reward structures were created to characterize preferred behavior. This information can be used to guide the PBN to avoid fault conditions and failures. (shrink)

Contemporary epistemological and cognitive studies, as well as recent trends in computer science and game theory have revealed an increasingly important and intimate relationship between Information, Interaction, and Agency. Agents perform actions based on the available information and in the presence of other interacting agents. From this perspective Information, Interaction, and Agency neatly ties together classical themes like rationality, decision-making and belief revision with games, strategies and learning in a multi-agent setting. Unified by the central notions Information, Interaction, and (...) Agency, the essays in this volume provide refreshing methodological perspectives on belief revision, dynamic epistemic logic, von Neumann games, and evolutionary game theory; all of which in turn are central approaches to understanding our own rationality and that of other agents. (shrink)

Typically, fair machine learning research focuses on a single decision maker and assumes that the underlying population is stationary. However, many of the critical domains motivating this work are characterized by competitive marketplaces with many decision makers. Realistically, we might expect only a subset of them to adopt any non-compulsory fairness-conscious policy, a situation that political philosophers call partial compliance. This possibility raises important questions: how does partial compliance and the consequent strategic behavior of decision subjects affect the allocation outcomes? (...) If k% of employers were to voluntarily adopt a fairness-promoting intervention, should we expect k% progress (in aggregate) towards the benefits of universal adoption, or will the dynamics of partial compliance wash out the hoped-for benefits? How might adopting a global (versus local) perspective impact the conclusions of an auditor? In this paper, we propose a simple model of an employment market, leveraging simulation as a tool to explore the impact of both interaction effects and incentive effects on outcomes and auditing metrics. Our key findings are that at equilibrium: (1) partial compliance by k% of employers can result in far less than proportional (k%) progress towards the full compliance outcomes; (2) the gap is more severe when fair employers match global (vs local) statistics; (3) choices of local vs global statistics can paint dramatically different pictures of the performance vis-a-vis fairness desiderata of compliant versus non-compliant employers; (4) partial compliance based on local parity measures can induce extreme segregation. Finally, we discuss implications for auditors and insights concerning the design of regulatory frameworks. (shrink)

We provide a computational model of semantic alignment among communicating agents constrained by social and cognitive pressures. We use our model to analyze the effects of social stratification and a local transmission bottleneck on the coordination of meaning in isolated dyads. The analysis suggests that the traditional approach to learning—understood as inferring prescribed meaning from observations—can be viewed as a special case of semantic alignment, manifesting itself in the behaviour of socially imbalanced dyads put under mild pressure of a local (...) transmission bottleneck. Other parametrizations of the model yield different long-term effects, including lack of convergence or convergence on simple meanings only. (shrink)

Résumé À partir de données expérimentales issues d’un jeu de la confiance répété, nous estimons des modèles structurels de formation des croyances permettant de distinguer les modes d’apprentissage des deux joueurs. Nous trouvons que les deux joueurs ne peuvent être décrits par le même mode d’apprentissage. Des simulations sur longue période montrent ensuite que l’interaction de ces deux types d’agents peut conduire à des issues contrastées.

Explanations of cognitive processes provided by traditional artificial intelligence were based on the notion of the knowledge level. This perspective has been challenged by new AI that proposes an approach based on embodied systems that interact with the real‐world. We demonstrate that these two views can be unified. Our argument is based on the assumption that knowledge level explanations can be defined in the context of Bayesian theory while the goals of new AI are captured by using a well established (...) robot based model of learning and problem solving, called Distributed Adaptive Control (DAC). In our analysis we consider random foraging and we prove that minor modifications of the DAC architecture renders a model that is equivalent to a Bayesian analysis of this task. Subsequently, we compare this enhanced, “rational,” model to its “non‐rational” predecessor and a further control condition using both simulated and real robots, in a variety of environments. Our results show that the changes made to the DAC architecture, in order to unify the perspectives of old and new AI, also lead to a significant improvement in random foraging. (shrink)

Explanations of cognitive processes provided by traditional artificial intelligence were based on the notion of the knowledge level. This perspective has been challenged by new AI that proposes an approach based on embodied systems that interact with the real‐world. We demonstrate that these two views can be unified. Our argument is based on the assumption that knowledge level explanations can be defined in the context of Bayesian theory while the goals of new AI are captured by using a well established (...) robot based model of learning and problem solving, called Distributed Adaptive Control (DAC). In our analysis we consider random foraging and we prove that minor modifications of the DAC architecture renders a model that is equivalent to a Bayesian analysis of this task. Subsequently, we compare this enhanced, “rational,” model to its “non‐rational” predecessor and a further control condition using both simulated and real robots, in a variety of environments. Our results show that the changes made to the DAC architecture, in order to unify the perspectives of old and new AI, also lead to a significant improvement in random foraging. (shrink)

We propose an experimental method to study the possible emergence of sensemaking in artificial agents. This method involves analyzing the agent's behavior in a test bed environment that presents regularities in the possibilities of interaction afforded to the agent, while the agent has no presuppositions about the underlying functioning of the environment that explains such regularities. We propose a particular environment that permits such an experiment, called the Small Loop Problem. We argue that the agent's behavior (...) demonstrates sensemaking if the agent learns to exploit regularities of interaction to fulfill its self-motivation as if it understood (at least partially) the underlying functioning of the environment. As a corollary, we argue that sensemaking and self-motivation come together. We propose a new method to generate self-motivation in an artificial agent called interactional motivation. An interactionally motivated agent seeks to perform interactions with predefined positive values and avoid.. (shrink)

This paper presents a human-computer interaction model with a three layers learning mechanism in a pervasive environment. We begin with a discussion around a number of important issues related to human-computer interaction followed by a description of the architecture for a multi-agent cooperative design system for pervasive computing environment. We present our proposed three- layer HCI model and introduce the group formation algorithm, which is predicated on a dynamic sharing niche technology. Finally, we explore the cooperative reinforcement learning and (...) fusion algorithms; the paper closes with concluding observations and a summary of the principal work and contributions of this paper. (shrink)

This article describes and evaluates an approach to create and/or improve tactical agents through direct human interaction in real time through a force-feedback haptic device. This concept takes advantage of a force-feedback joystick to enhance motor skill and decision-making transfer from the human to the agent in real time. Haptic devices have been shown to have high bandwidth and sensitivity. Experiments are described for this new approach, named Instructional Learning. It is used both as a way to build agents (...) from scratch as well as to improve and/or correct agents built through other means. The approach is evaluated through experiments that involve three applications of increasing complexity – chasing a fleer, shepherding a flock of sheep into a pen, and driving a virtual automobile through a simulated road network. The results indicate that in some instances, instructional learning can successfully create agents under some circumstances. However, instructional learning failed to build and/or improve agents in other instances. The Instructional Learning approach, the experiments, and their results are described and extensively discussed. (shrink)

In this paper, we model a relational notion of subjectivity by means of two experiments in subjective computing. The goal is to determine to what extent a cognitive and social robot can be regarded to act subjectively. The system was implemented as a reinforcement learning agent with a coaching function. To analyze the robotic agent we used the method of levels of abstraction in order to analyze the agent at four levels of abstraction. At one level the (...) agent is described in mentalistic or subjective language respectively. By mapping this mentalistic to an algorithmic, functional, and relational level, we can show to what extent the agent behaves subjectively as we make use of a relational concept of subjectivity that draws upon the relations that hold between the agent and its environment. According to a relational notion of subjectivity, an agent is supposed to be subjective if it exhibits autonomous relations to itself and others, i.e. the agent is not fully determined by a given input but is able to operate on its input and decide what to do with it. This theoretical notion is confirmed by the technical implementation of self-referentiality and social interaction in that the agent shows improved behavior compared to agents without the ability of subjective computing. On the one hand, a relational concept of subjectivity is confirmed, whereas on the other hand, the technical framework of subjective computing is being theoretically founded. (shrink)

This paper introduces a hybrid model that unifies connectionist, symbolic, and reinforcement learning into an integrated architecture for bottom-up skill learning in reactive sequential decision tasks. The model is designed for an agent to learn continuously from on-going experience in the world, without the use of preconceived concepts and knowledge. Both procedural skills and high-level knowledge are acquired through an agent’s experience interacting with the world. Computational experiments with the model in two domains are reported.

The mechanistic model depicts scientific explanations as involving the discovery of multi-level, organized components that constitute a target phenomenon. Meanwhile, sensorimotor enactivism purports to offer a scientifically informed account of perceptual experience as a skill-laden interactive relationship, constitutively involving both perceiver and world, rather than as an agent-bound representation of the world. Insofar as sensorimotor enactivism identifies an empirically tractable phenomenon – skillful agent-world interaction – and mechanistic explanation establishes the subpersonal components of this phenomenon, the two approaches (...) allow for a fruitful division of labor in investigating perceptual experience. On closer inspection, however, two challenges arise. First, the “representation challenge” arises because promising attempts to set out implementational details of our sensorimotor interaction with the world implicate cognitive representations, creating tension with sensorimotor enactivism’s nonrepresentational commitments. Second, the “reconstitution challenge” arises when mechanistic explanation not only uncovers the components of some established phenomenon but plays a role in “reconstituting” this phenomenon. This means that, through investigating mechanisms, perceptual experience may be reconceived such that its constituents are wholly organism-bound. We explore both challenges to the compatibility of mechanism and sensorimotor enactivism and examine possible solutions. The result is a clearer understanding of the tensions and opportunities for learning between the frameworks. (shrink)

This comment discusses work by Aylett and Paiva (2012) which describes a synthetic approach to building a virtual world inhabited by synthetic characters where the user can experience subjective culture, that is, the experience of social reality, and learn how to empathetically communicate with people in other cultures. It provides a computational theory for integrating recent findings on emotion and cultural sensitivities into an interactive drama played by interacting characters with varying personalities. The FAtiMA-PSI, the implementation of their theory, has (...) advanced a critical step towards basing affective intelligence on a firmer ground. Future challenges include building an empathic agent on the simulated physiology level that can be recognized as an independent ego with its own personality and presence. (shrink)

Artificial agents are progressively becoming more present in everyday-life situations and more sophisticated in their interaction affordances. In some specific cases, like Google Duplex, GPT-3 bots or Deep Mind’s AlphaGo Zero, their capabilities reach or exceed human levels. The use contexts of everyday life necessitate making such agents understandable by laypeople. At the same time, displaying human levels of social behavior has kindled the debate over the adoption of Dennett’s ‘intentional stance’. By means of a comparative analysis of the literature (...) on robots and virtual agents, we defend the thesis that approaching these artificial agents ‘as if’ they had intentions and forms of social, goal-oriented rationality is the only way to deal with their complexity on a daily base. Specifically, we claim that this is the only viable strategy for non-expert users to understand, predict and perhaps learn from artificial agents’ behavior in everyday social contexts. Furthermore, we argue that as long as agents are transparent about their design principles and functionality, attributing intentions to their actions is not only essential, but also ethical. Additionally, we propose design guidelines inspired by the debate over the adoption of the intentional stance. (shrink)

A shared narrative in the literature on the evolution of cooperation maintains that social _learning_ evolves early to allow for the transmission of cumulative culture. Social _norms_, whilst present at the outset, only rise to prominence later on, mainly to stabilise cooperation against the threat of defection. In contrast, I argue that once we consider insights from social epistemology, an expansion of this narrative presents itself: An interesting kind of social norm — an epistemic coordination norm — was operative in (...) early and important instances of specialised social learning. I show how there’s a need for such norms in two key social learning strategies and explain how this need is constituted. In assessor-teaching (e.g. Castro et al., 2019b, 2021), epistemic coordination norms allow agents to coordinate around the _content_ of social learning, i.e., what is to be known and how this is to be done. These norms also allow agents to coordinate around the form of cultural learning in what’s sometimes called strategic social learning (Laland, 2004; Hoppitt & Laland, 2013; Heyes, 2018, Chap. 5) and elsewhere. Broadly speaking, this concerns how cultural learning is organised within the social group. The upshot is that the evolution of social learning and social norms are intertwined in important and underappreciated ways from early on. The above matters as it informs our views about the evolution of social norms more generally. Truly _social_ norms emerged to coordinate a plurality of complex behaviours and interactions, amongst them specialised social learning. I substantiate this view by contrasting it with Jonathan Birch’s views on the evolution of norms. What results is a general but cohesive narrative on the early evolution of social norms. (shrink)

Scenarios for the emergence or bootstrap of a lexicon involve the repeated interaction between at least two agents who must reach a consensus on how to name N objects using H words. Here we consider minimal models of two types of learning algorithms: cross-situational learning, in which the individuals determine the meaning of a word by looking for something in common across all observed uses of that word, and supervised operant conditioning learning, in which there is strong feedback between individuals (...) about the intended meaning of the words. Despite the stark differences between these learning schemes, we show that they yield the same communication accuracy in the limits of large N and H , which coincides with the result of the classical occupancy problem of randomly assigning N objects to H words. Key words: lexicon bootstrapping; cross-situational learning; supervised learning; random occupancy problems. (shrink)

Scenarios for the emergence or bootstrap of a lexicon involve the repeated interaction between at least two agents who must reach a consensus on how to name N objects using H words. Here we consider minimal models of two types of learning algorithms: cross-situational learning, in which the individuals determine the meaning of a word by looking for something in common across all observed uses of that word, and supervised operant conditioning learning, in which there is strong feedback between individuals (...) about the intended meaning of the words. Despite the stark differences between these learning schemes, we show that they yield the same communication accuracy in the limits of large N and H, which coincides with the result of the classical occupancy problem of randomly assigning N objects to H words. Key words: lexicon bootstrapping; cross-situational learning; supervised learning; random occupancy problems. (shrink)

The explosive growth in computational power over the past several decades offers new tools and opportunities for economists. This handbook volume surveys recent research on Agent-based Computational Economics (ACE), the computational study of economic processes modeled as open-ended dynamic systems of interacting agents. Empirical referents for “agents” in ACE models can range from individuals or social groups with learning capabilities to physical world features with no cognitive function. Topics covered include: learning; empirical validation; network economics; social dynamics; financial markets; (...) innovation and technological change; organizations; market design; automated markets and trading agents; political economy; social-ecological systems; computational laboratory development; and general methodological issues. (shrink)

The theme for this volume is social cognition, construed from a psychological and collective point of view. From the psychological point of view, the question is to understand how the human mind processes social information; how it encodes, stores and uses it in the social context. From a collective point of view, the question is to understand how individual cognition is influenced (improved, increased or impaired) by social interactions, for instance in communicating and collaborating with intelligent agents. These two dimensions (...) of social cognition are obviously interdependent: the psychological dimension makes the collective dimension possible, which can in return modify the psychological dimension. The book is divided into four parts. The first part is about socio-cognitive skills. Among those, we count face recognition, imitation learning, embodied social interaction, cheater detection and psychological concept acquisition. The second part is about persons and memories: stereotypes, attraction judgements and impression formation are the subjects at hand. The third part is about understanding each other. A key part of that understanding is the motor system (whether or not we see it as a "mirror"), but community membership itself can also contribute to our understanding of others. The fourth and final part is about social cognition in societies. This section is unified by the common goal of understand how social cognition actually influences the structure of different societies, whether whole cultures, specific social networks, rural communities or even groups of caterpillars! (shrink)

In this paper, we propose a learning method for an agent to interact with other agents effectively. This method, MULA-C, improves efficiency of the learning, by clustering agents, and influences the learning experience of one agent to other agents which belong to the same cluster. Similarity among agents is evaluated by similarity among Q-values of agents. We give the detail explanation of learning method of MULA-C, and present the result of experiments which shows the effectiveness of MULA-C.

Literature on consumers’ ethical decision making is rooted in a rationalist perspective that emphasizes the role of moral reasoning. However, the view of ethical consumption as a thorough rational and conscious process fails to capture important elements of human cognition, such as emotions and intuitions. Based on moral psychology and microsociology, this paper proposes a holistic and integrated framework showing how emotive and intuitive information processing may foster ethical consumption at individual and social levels. The model builds on social intuitionism (...) to show how consumers’ a priori affect-laden intuitive moral judgments impact their post hoc reflective moral reasoning. Symbolic interactionism is used to interpret consumers as interdependent and socially embedded agents that self-construct their social identity through interactions with other consumers. The proposed social intuitionist framework of consumers’ ethical decision making shows that other-oriented moral emotions—such as elevation, gratitude, and empathy—interact with persuasion and social influence in ethical consumption. Consequently, moral emotions and intuition drive interpersonal persuasion among ethical consumers. Theoretical propositions and implications for consumer ethics theory and practice are discussed. (shrink)

The article focuses on the study of workplace learning for hospital’s staff to become mortuary care agents. It is based on the analysis of verbalizations obtained from self-confrontations interviews made with the mortuary care trainees. These interviews are extracted from video recordings taken while trainees carry out their normal task. The goal is to identify, within the context of the interaction between trainees and tutors, signs of learning progress in the trainees’ verbalizations, which could be linked to existing hospital standards. (...) As demonstrated by the activity clinic methodology, these verbalizations are not just a simple depiction of the work activity as it happened but serve as an experiment to underline the learning benefits for the trainees. Beyond the benefits of better understanding the trainees work tasks, the resulting analyses indeed support the formative aspect of self-confrontation interviews. In this perspective, the article focuses on a methodological aspect, the data analysis to account for signs of learning such as verbalized and recognized by the trainees in their interactions with tutors. On the occasion, we question the formative dimension of this methodological setting in the prospect of designing training devices for these professionals. (shrink)