Cognitive Systems Research最新文献

Despite lacking a generally accepted definition, Artificial General Intelligence (AGI) is commonly understood to refer to artificial agents possessing the capacity to build up a context-independent understanding of itself and the world and to generalize this knowledge across a multitude of contexts. In human agents, this capacity is, to a large degree, facilitated by processes of self-directed learning, during which agents voluntarily control the conditions under which episodes of learning and problem solving occur. Since self-directed learning depends on the degree of knowledge the agent has about various aspects of themselves (their bodily skills, their learning goal, etc.), an AGI implementation of this type of learning must build on a theory of how this self-knowledge is actualized and modified during the learning process. In this paper, we employ the pattern theory of self in order to characterize different aspects of an agent’s self that are relevant for self-directed learning. Such aspects include agent-internal cognitive states such as thoughts, emotions, and intentions, but also relational states such as action possibilities in the environment. Combinations of these aspects form a characteristic pattern, which is unique to each individual agent, with no one aspect being necessary or sufficient for the individuation of that agent’s self. Here, we focus on the interdependence of narrative and embodied aspects of the self-pattern, since they involve particularly salient challenges consisting in conceptualizing the interaction between propositional and motor representations.

In our paper, we model the reciprocal interaction of these aspects of the self-pattern within an individual cognitive agent. We do so by extending an approach by Ryan, Agrawal, & Franklin (2020), who laid the groundwork for the implementation of the pattern theory of self in the LIDA (Learning Intelligent Decision Agent) model. We describe how embodied and narrative aspects of an agent’s self-pattern are realized by patterns of interaction between different LIDA modules over time, and how interactions at multiple temporal scales allow the agent’s self-pattern to be both dynamically variable and relatively stable. Finally, we investigate the implications this view has for the creation of artificial agents that can benefit from self-directed learning, both in the context of deliberate planning and adaptive motor execution.

Advancements in autonomy are leading to an increased need for machines capable of collaborative effort with humans to achieve team goals. One way of enhancing these human-autonomous system work arrangements leverages the concept of a shared mental model. The idea being that when the human and autonomous teammate have aligned models, the team is more productive due to an increase in trust, predictiveness, and apparent understanding. An open issue is how to have autonomous teammates learn a user aligned mental model. This research presents a dual-process learning model that leverages multivariate normal probability density functions (DPL-MN) to extrapolate state-responses into system 2. By leveraging dual-process learning concepts, an autonomous teammate is able to rapidly align with a user and extrapolate their consistencies into longer term memory. Evaluation of DPLM with user responses from a game called Space Navigator shows that DPL-MN accurately responds to situations similarly to each unique user.

Collaborative teams pursue common goals, completing tasks and making decisions with various levels of interdependence. A shared mental model (SMM) is a foundational structure in high performing, production teams and aids humans in predicting their teammate’s goals and intentions. Advice teams utilize a transactive memory system (TMS) that integrate sources of knowledge and the source’s credibility. SMMs and TMSs elevate human performance when the nature of emergence complements the associated team type. However, project and action teams require both behavioral and knowledge integration. We present a hybrid cognitive model (HCM) for machine agents that unifies SMM and TMS characteristics. The HCM enables anytime selection over the two cognitive representations with the computational complexity of a single model. Furthermore, the pliant nature of credibility modeling in TMSs can represent expertise, thoroughness, or trust simultaneously in the team context. Results in a multi-agent project domain demonstrate the HCM’s efficacy for machine agent teams and potential for applications in human–machine teams.

The paper considers the task-driven approach to artificial intelligence. It is shown that, on the one hand, it generalizes such approaches as the agent-based approach and general artificial intelligence, and, on the other hand, accurately reflects the cognitive processes and purposeful behavior described in the physiological Theory of functional brain systems.

The paper presents repeatable effects of synchronizing visual and auditory alarms with heartbeats on the availability of cognitive resources. A perception-driven situation awareness model is proposed and studied by implementing two distinct experimental protocols with different groups of participants. Results of a first study with a single-screen configuration are repeated by those of a second one on a multiple-screen context. Both experimental protocols rely on manipulating a between-subjects factor to compare two conditions - one with alarms activated synchronously with heart rate and one with alarms non-synchronized with heart rate - and a within-subjects factor to compare the impact of workload by increasing the level of task difficulty. Results about mono-screen and multi-screen configurations are homogenous. The synchronous condition makes people produce significantly more errors and fewer visual scans of the alarm display area. This degradation of perceptual abilities is non-conscious and is correlated with workload. Main people are not aware about their actual performance and this is confirmed by the evolution of subjective performance and frustration regarding task difficulty, display configuration and alarm activation condition. Such discrepancies between what it is looked at with what it is actually perceived and between actual and perceived indicators like performance are perceptual dissonances that are relevant for perception-driven situation awareness. The application of synchronizing dynamic events with heartbeats will be studied for different individual and collective work contexts in order to extend the proposed perception-driven situation awareness model based on perceptual dissonance management and on human capability parameters.

This study proposes a model of computational consciousness for non-interacting agents. The phenomenon of interest was assumed as sequentially dependent on the cognitive tasks of sensation, perception, emotion, affection, attention, awareness, and consciousness. Starting from the Smart Sensing prodromal study, the cognitive layers associated with the processes of attention, awareness, and consciousness were formally defined and tested together with the other processes concerning sensation, perception, emotion, and affection. The output of the model consists of an index that synthesizes the energetic and entropic contributions of consciousness from a computationally moral perspective. Attention was modeled through a bottom-up approach, while awareness and consciousness by distinguishing environment from subjective cognitive processes. By testing the solution on visual stimuli eliciting the emotions of happiness, anger, fear, surprise, contempt, sadness, disgust, and the neutral state, it was found that the proposed model is concordant with the scientific evidence concerning covert attention. Comparable results were also obtained regarding studies investigating awareness as a consequence of visual stimuli repetition, as well as those investigating moral judgments to visual stimuli eliciting disgust and sadness. The solution represents a novel approach for defining computational consciousness through artificial emotional activity and morality.

Previous reports show that a substantial proportion of (near) medical errors in the operating theatre is attributable to ineffective communication between healthcare professionals. Speaking up about observed medical errors is a safety behaviour which promotes effective communication between health care professionals, consequently optimising patient care by reducing medical error risk. Speaking up by healthcare professionals (e.g., nurses, residents) remains difficult to execute in practice despite increasing awareness of its importance. Therefore, this paper discourses a computational model concerning the mechanisms known from psychological, observational, and medical literature which underlie the speaking up behaviour of a health care professional. It also addresses how a doctor may respond to the communicated message. Through several scenarios we illustrate what pattern of factors causes a healthcare professional to speak up when witnessing a (near) medical error. We moreover demonstrate how introducing an observant agent can facilitate effective communication and help to ensure patient safety through speaking up when a nurse can not. In conclusion, the current paper introduces an adaptive computational model which predicts speaking up behaviour from the perspective of the speaker and receiver, with the addition of a virtual coach to further optimise patient safety when a patient could be in harm’s way.

The current success of Reinforcement Learning algorithms for its performance in complex environments has inspired many recent theoretical approaches to cognitive science. Artistic environments are studied within the cognitive science community as rich, natural, multi-sensory, multi-cultural environments. In this work, we propose the introduction of Reinforcement Learning for improving the control of artistic robot applications. Deep Q-learning Neural Networks (DQN) is one of the most successful algorithms for the implementation of Reinforcement Learning in robotics. DQN methods generate complex control policies for the execution of complex robot applications in a wide set of environments. Current art painting robot applications use simple control laws that limits the adaptability of the frameworks to a set of simple environments. In this work, the introduction of DQN within an art painting robot application is proposed. The goal is to study how the introduction of a complex control policy impacts the performance of a basic art painting robot application. The main expected contribution of this work is to serve as a first baseline for future works introducing DQN methods for complex art painting robot frameworks. Experiments consist of real world executions of human drawn sketches using the DQN generated policy and TEO, the humanoid robot. Results are compared in terms of similarity and obtained reward with respect to the reference inputs.

People segment complex, ever-changing, and continuous experience into basic, stable, and discrete spatio-temporal experience units, called events. The literature on event segmentation investigates the mechanisms behind this ability. Event segmentation theory points out that people predict ongoing activities and observe prediction error signals to find event boundaries. In this study, we investigated the mechanism giving rise to this ability through a computational model and accompanying psychological experiments. Inspired by event segmentation theory and predictive processing, we introduced a self-supervised model of event segmentation. This model consists of neural networks that predict the sensory signal in the next time-step to represent different events, and a cognitive model that regulates these networks on the basis of their prediction errors. In order to verify the ability of our model in segmenting events, learning them during passive observation, and representing them in its representational space, we prepared a video of human behaviors represented by point-light displays. We compared the event segmentation behaviors of participants and our model with this video in two granularities. Using point-biserial correlation, we demonstrated that the event boundaries of our model correlated with the responses of the participants. Moreover, by approximating the representation space of participants, we showed that our model formed a similar representation space with those of participants. The result suggests that our model that tracks the prediction error signals can produce human-like event boundaries and event representations. Finally, we discuss our contribution to the literature and our understanding of how event segmentation is implemented in the brain.