Goffin's cockatoos () can solve a diverse set of mechanical problems, such as tool use, tool manufacture, and mechanical puzzles. However, the proximate mechanisms underlying this adaptive behavior are largely unknown. Similarly, engineering artificial agents that can as flexibly solve such mechanical puzzles is still a substantial challenge in areas such as robotics. This article is an interdisciplinary approach to study mechanical problem solving which we hope is relevant to both fields. The behavior we are studying results from the interaction between a complex environment (the lockbox) and different processes that govern the animals' behavior. We therefore jointly analyze the parrots' (1) engagement, (2) sensorimotor skill learning, and (3) action selection. We find that none of these aspects could solely explain the animals' behavioral adaptation and that a plausible model of proximate mechanisms must jointly address these aspects. We accompany this analysis with a discussion of methods to identify such mechanisms. At the same time, we argue, it is implausible to identify a detailed model from the limited behavioral data of just a few studies. Instead, we advocate for an incremental approach to model building in which one first establishes constraints on proximate mechanisms before specific, detailed models are formulated. To illustrate this idea, we apply it to the data presented here. We argue that as the field attempts to find mechanistic explanations for increasingly complex behaviors, such alternative modeling approaches will be necessary.

An affordance perspective highlights how resourceful the ecology is for creative actions of all sorts; it captures how creativity is grounded in materiality. In contrast to "canonical affordances" (i.e., "ready-to-hand," mundane instances), creative affordances point to unconventional or surprising action opportunities that are nonetheless valued. Our initial aim is to discuss how to frame the affordance concept to make it attractive for the study of creativity. We propose a dialectic position that reconciles aspects of the realism of ecological psychology with the constructivist view more typical of creativity scholars. We stress that novel options frequently depend on constructive actions; novelty cannot always simply be "found" or just waits to be used. Many creative opportunities only emerge from how person actively engages with the ecology. Our second aim is to explore specific ways that creativity is mediated through affordances, based on illustrations from crafts and dance. These suggest that affordances span various timescales and mediate in multiple ways, from noticing existing potentials, via active affordance shaping, to background activities that indirectly invite or enable novelty. In conclusion we discuss how a person's creative "vision," imagination and combinatoric ability, all fundamental creativity mechanisms, relate to affordances and how fruitful creative directions may be perceptually hinted at.

Constructing general knowledge by learning task-independent models of the world can help agents solve challenging problems. However, both constructing and evaluating such models remain an open challenge. The most common approaches to evaluating models is to assess their accuracy with respect to observable values. However, the prevailing reliance on estimator accuracy as a proxy for the usefulness of the knowledge has the potential to lead us astray. We demonstrate the conflict between accuracy and usefulness through a series of illustrative examples including both a thought experiment and an empirical example in Minecraft, using the General Value Function framework (GVF). Having identified challenges in assessing an agent's knowledge, we propose an alternate evaluation approach that arises naturally in the online continual learning setting: we recommend evaluation by examining internal learning processes, specifically the relevance of a GVF's features to the prediction task at hand. This paper contributes a first look into evaluation of predictions through their use, an integral component of predictive knowledge which is as of yet unexplored.

We present three new diagnostic prediction problems inspired by classical-conditioning experiments to facilitate research in online prediction learning. Experiments in classical conditioning show that animals such as rabbits, pigeons, and dogs can make long temporal associations that enable multi-step prediction. To replicate this remarkable ability, an agent must construct an internal state representation that summarizes its interaction history. Recurrent neural networks can automatically construct state and learn temporal associations. However, the current training methods are prohibitively expensive for -continual learning on every time step-which is the focus of this paper. Our proposed problems test the learning capabilities that animals readily exhibit and highlight the limitations of the current recurrent learning methods. While the proposed problems are nontrivial, they are still amenable to extensive testing and analysis in the small-compute regime, thereby enabling researchers to study issues in isolation, ultimately accelerating progress towards scalable online representation learning methods.

With this inaugural lecture as Socrates Professor on the topic of Making Humane Technologies, I aim to show that artistic practices afford embedding technologies better in society. Analyzing artworks made at RAAAF, an art collective that makes visual art and experimental architecture, I will describe three aspects of making practices that may contribute to improving the embedding of technology in society: (1) the skill of working with layers of meaning; (2) the skill of creating material playgrounds that afford free exploration of the potential of new technologies and artistic experiments; and (3) the skill of openness to the possibility of having radically different socio-material practices. I will use images of several RAAAF projects to make these skills involved in making more tangible. It is artistic skills like these that can contribute to a better societal embedding of technologies.

The aim of this article is to clarify how best to interpret some of the central constructs that underwrite the free-energy principle (FEP) - and its corollary, active inference - in theoretical neuroscience and biology: namely, the role that generative models and variational densities play in this theory. We argue that these constructs have been systematically misrepresented in the literature, because of the conflation between the FEP and active inference, on the one hand, and distinct (albeit closely related) Bayesian formulations, centred on the brain - variously known as predictive processing, predictive coding or the prediction error minimisation framework. More specifically, we examine two contrasting interpretations of these models: a structural representationalist interpretation and an enactive interpretation. We argue that the structural representationalist interpretation of generative and recognition models does not do justice to the role that these constructs play in active inference under the FEP. We propose an enactive interpretation of active inference - what might be called . In active inference under the FEP, the generative and recognition models are best cast as realising inference and control - the self-organising, belief-guided selection of action policies - and do not have the properties ascribed by structural representationalists.

Enactive approaches to cognitive science aim to explain human cognitive processes across the board without making any appeal to internal, content-carrying representational states. A challenge to such a research programme in cognitive science that immediately arises is how to explain cognition in so-called 'representation-hungry' domains. Examples of representation-hungry domains include imagination, memory, planning and language use in which the agent is engaged in thinking about something that may be absent, possible or abstract. The challenge is to explain how someone could think about things that are not concretely present in their environment other than by means of an internal mental representation. We call this the 'Representation-Hungry Challenge' (RHC). The challenge we take up in this article is to show how hunger for representations could possibly be satisfied by means other than the construction and manipulation of internal representational states. We meet this challenge by developing a theoretical framework that integrates key ideas drawn from enactive cognitive science and ecological psychology. One of our main aims is thus to show how ecological and enactive theories as non-representational and non-computational approaches to cognitive science might work together. From enactive cognitive science, we borrow the thesis of the strict continuity of lower and higher cognition. We develop this thesis to argue against any sharp conceptual distinction between higher and lower cognition based on representation-hunger. From ecological psychology, we draw upon our earlier work on the rich landscape of affordances. We propose thinking of so-called representation-hungry cognition in terms of temporally extended activities in which the agent skilfully coordinates to a richly structured landscape of affordances. In our framework, putative cases of representation-hungry cognition are explained by abilities to coordinate nested activities to an environment structured by interrelated socio-material practices. The RHC has often figured in arguments for the limitations of non-representational approaches to cognitive science. We showcase the theoretical resources available to an integrated ecological-enactive approach for addressing this type of sceptical challenge.

In this paper, the role of adaptive group cohesion in a cooperative multi-agent source localization problem is investigated. A distributed source localization algorithm is presented for a homogeneous team of simple agents. An agent uses a single sensor to sense the gradient and two sensors to sense its neighbors. The algorithm is a set of individualistic and social behaviors where the individualistic behavior is as simple as an agent keeping its previous heading and is not self-sufficient in localizing the source. Source localization is achieved as an emergent property through agent's adaptive interactions with the neighbors and the environment. Given a single agent is incapable of localizing the source, maintaining team connectivity at all times is crucial. Two simple temporal sampling behaviors, intensity-based-adaptation and connectivity-based-adaptation, ensure an efficient localization strategy with minimal agent breakaways. The agent behaviors are simultaneously optimized using a two phase evolutionary optimization process. The optimized behaviors are estimated with analytical models and the resulting collective behavior is validated against the agent's sensor and actuator noise, strong multi-path interference due to environment variability, initialization distance sensitivity and loss of source signal.

A robot capable of understanding emotion expressions can increase its own capability of solving problems by using emotion expressions as part of its own decision-making, in a similar way to humans. Evidence shows that the perception of human interaction starts with an innate perception mechanism, where the interaction between different entities is perceived and categorized into two very clear directions: positive or negative. While the person is developing during childhood, the perception evolves and is shaped based on the observation of human interaction, creating the capability to learn different categories of expressions. In the context of human-robot interaction, we propose a model that simulates the innate perception of audio-visual emotion expressions with deep neural networks, that learns new expressions by categorizing them into emotional clusters with a self-organizing layer. The proposed model is evaluated with three different corpora: The Surrey Audio-Visual Expressed Emotion (SAVEE) database, the visual Bi-modal Face and Body benchmark (FABO) database, and the multimodal corpus of the Emotion Recognition in the Wild (EmotiW) challenge. We use these corpora to evaluate the performance of the model to recognize emotional expressions, and compare it to state-of-the-art research.

We present a robot architecture and experiments to investigate some of the roles that pleasure plays in the decision making (action selection) process of an autonomous robot that must survive in its environment. We have conducted three sets of experiments to assess the effect of different types of pleasure-related versus unrelated to the satisfaction of physiological needs-under different environmental circumstances. Our results indicate that pleasure, including pleasure unrelated to need satisfaction, has value for homeostatic management in terms of improved viability and increased flexibility in adaptive behavior.

Consensus formation is investigated for multi-agent systems in which agents' beliefs are both vague and uncertain. Vagueness is represented by a third truth state meaning . This is combined with a probabilistic model of uncertainty. A belief combination operator is then proposed, which exploits borderline truth values to enable agents with conflicting beliefs to reach a compromise. A number of simulation experiments are carried out, in which agents apply this operator in pairwise interactions, under the bounded confidence restriction that the two agents' beliefs must be sufficiently consistent with each other before agreement can be reached. As well as studying the consensus operator in isolation, we also investigate scenarios in which agents are influenced either directly or indirectly by the state of the world. For the former, we conduct simulations that combine consensus formation with belief updating based on evidence. For the latter, we investigate the effect of assuming that the closer an agent's beliefs are to the truth the more visible they are in the consensus building process. In all cases, applying the consensus operators results in the population converging to a single shared belief that is both crisp and certain. Furthermore, simulations that combine consensus formation with evidential updating converge more quickly to a shared opinion, which is closer to the actual state of the world than those in which beliefs are only changed as a result of directly receiving new evidence. Finally, if agent interactions are guided by belief quality measured as similarity to the true state of the world, then applying the consensus operator alone results in the population converging to a high-quality shared belief.

A stimulus is a reliable signal of an outcome when the probability that the outcome occurs in its presence is different from in its absence. Reliable signals of important outcomes are responsible for triggering critical anticipatory or preparatory behavior, which is any form of behavior that prepares the organism to receive a biologically significant event. Previous research has shown that humans and other animals prepare more for outcomes that occur in the presence of highly reliable (i.e., highly contingent) signals, that is, those for which that difference is larger. However, it seems reasonable to expect that, all other things being equal, the probability with which the outcome follows the signal should also affect preparatory behavior. In the present experiment with humans, we used two signals. They were differentially followed by the outcome, but they were equally (and relatively weakly) reliable. The dependent variable was preparatory behavior in a Martians video game. Participants prepared more for the outcome (a Martians' invasion) when the outcome was most probable. These results indicate that the probability of the outcome can bias preparatory behavior to occur with different intensities despite identical outcome signaling.

The agent-based model presented here builds on existing models, allowing for multiple partnerships, including those overlapping in time, to examine sexual partnerships, with the goal of hypothesis testing and guiding data collection. Within each model run, agents are assigned characteristics (including quality, aspiration, courtship duration, and ideal number of lifetime partners) and then search for partners; existing couples choose whether they should break up, remain dating, or become sexual partners. Model behavior was tested across a wide range of parameters and compared with empirical data. The model produces numbers of lifetime sexual partners, and partners in the last year, rates of concurrency, and relationship durations similar to nationally representative data from the US; it also generates correlations in partners' quality similar to those reported for marriage and dating partners. Model results highlight the importance of individual preferences, interactions between individuals, and contextual factors in sexual decision-making.

The Iterated Classification Game (ICG) combines the Classification Game with the Iterated Learning Model (ILM) to create a more realistic model of the cultural transmission of language through generations. It includes both learning from parents and learning from peers. Further, it eliminates some of the chief criticisms of the ILM: that it does not study grounded languages, that it does not include peer learning, and that it builds in a bias for compositional languages. We show that, over the span of a few generations, a stable linguistic system emerges that can be acquired very quickly by each generation, is compositional, and helps the agents to solve the classification problem with which they are faced. The ICG also leads to a different interpretation of the language acquisition process. It suggests that the role of parents is to initialize the linguistic system of the child in such a way that subsequent interaction with peers results in rapid convergence to the correct language.