Cognitive Architectures

In what concerns living systems, cognition is an embodied, embedded and always situated experience. This means that it involves an entity endowed with a particular physical architecture bound in a dialectical relationship with the environment in which it is immersed, behaving according to the prompts placed by this environment, reacting, learning and adapting to it defining this way its own existential narrative and history. Highlighting the fact that human cognition stems from more simple and basic forms of cognition with which it shares essential life mechanisms, the present chapter focuses on the essential semiosic process that is inherent to the dialectics agent/environment and the role played by corporeal architectures in the construction of meaningful worlds, namely, the hybrid realities, where natural and artificial intelligence cohabit.

In this chapter, we use the example of colouration of animal surfaces to show how processes based on interactions of the individual parts enter, as units, into processes on other levels and how this processual scaffolding leads to the emergence of ‘meaning’ on the level of communication between individuals. We review recent understanding of colour production and pattern formation in animals. We describe self-organization and dynamical nature of these processes. To highlight the inseparability of seeing and appearing, we discuss shared evolutionary origins of sight and colouration. Common evolutionary explanations of colouration are then discussed. Due to the complementarity of appearance and perception, the exposed surfaces of organisms ultimately become semi-autonomous entities subjected to their own evolution.

We extend our physical selfs, perceptual and cognitive realities as well as memories and imagination through countless technical inventions and conceptual systems. In his book The Extended Phenotype, the biologist Richard Dawkins, suggests that in the biological world such extensions are so important that, for instance, the dams and water regulation systems of the beaver should be included in the biological definition of the species of the beaver. Similarly, our countless constructions, structures, technical systems as well as intellectual discoveries, ought to be included in the definition of Homo Sapiens, but we still continue to see ourselves limited by our skin. Altogether, we tend to think of our environments in terms of isolated, definable objects and entities, rather than dynamic and constantly interactive and expanding systems. Architecture is likewise seen as material aestheticized structures that are external to us, rather than as part of our biological and mental constitution. However, our environments from intimate objects to rooms, buildings, cities, regions and all the way to the entire world and the universe, can also be regarded as part of our material, perceptual, and conceptual reality. Instead of being seen as material objects and buildings, architecture should be regarded as an active entity which very concretely mediates our relationships with the world through space and time. Human history, culture, and collective consciousness widen our world of thought and action beyond material boundaries. Through our structures, we, humans, turn limitless, shapeless and meaningless space into lived space with human meanings. We also regard architecture as an aesthetic expression of its architect, but Maurice Merleau-Ponty argues thought provokingly: “We come to see not the work of art, but the world according to the work”. Architecture has a crucial role in the constitution of the human world, both material and mental.

Given that original purpose of cognitive architectures was to lead to a unified theory of cognition, this chapter considers the possible contributions that cognitive architectures can make to embodied theories of cognition in particular. This is not a trivial question since the field remains very much divided about what embodied cognition actually means, and we will see some example positions in this chapter. It is then argued that a useful embodied cognitive architecture would be one that can demonstrate (a) what precisely the role of the body in cognition actually is, and (b) whether a body is constitutively needed at all for some (or all) cognitive processes. It is proposed that such questions can be investigated if the cognitive architecture is designed so that consequences of varying the precise embodiment on higher cognitive mechanisms can be explored. This is in contrast with, for example, those cognitive architectures in robotics that are designed for specific bodies first; or architectures in cognitive science that implement embodiment as an add-on to an existing framework (because then, that framework is by definition not constitutively shaped by the embodiment). The chapter concludes that the so-called semantic pointer architecture by Eliasmith and colleagues may be one framework that satisfies our desiderata and may be well-suited for studying theories of embodied cognition further.

The creation of a cognitive architecture presents the architect with many design choices. Some of these choices come in the form of a dilemma, in which the selection of any option over another entails both benefits and opportunity costs. This chapter highlights three dilemmas that confront the architect when deciding how the key issues of fidelity, embodiment, and autonomy should be addressed and reflected in the design. In each case, it discusses the various options, their roots, and the consequences and costs of choosing one option over another. It concludes by considering these three dilemmas in the context of the stance on cognitive adopted by the editors of this book.

The hand is one of the most complex and fascinating organs of the human body. We can powerfully squeeze objects, but we are also capable of manipulating them with great precision and dexterity. On the other hand, the arm, with its redundant joints, is in charge of reaching the object by determining the hand pose during preshaping. The complex motion and task execution of the upper-limb system may lead us to think that the control requires a very significant brain effort. As a matter of fact, neuroscience studies demonstrate that humans simplify planning and control using a combination of primitives, which the brain modulates to produce hand configurations and force patterns for the purpose of grasping and manipulating different objects. This concept can be transferred to robotic systems, allowing control within a space of lower dimension. The lower number of parameters characterizing the system allows for embodying the control in machine learning frameworks, reproducing a sort of human-like cognition.

Synthetic psychology describes the approach of “understanding through building” applied to the human condition. In this chapter, we consider the specific challenge of synthesizing a robot “sense of self”. Our starting hypothesis is that the human self is brought into being by the activity of a set of transient self-processes instantiated by the brain and body. We propose that we can synthesize a robot self by developing equivalent sub-systems within an integrated biomimetic cognitive architecture for a humanoid robot. We begin the chapter by motivating this work in the context of the criteria for recognizing other minds, and the challenge of benchmarking artificial intelligence against human, and conclude by describing efforts to create a sense of self for the iCub humanoid robot that has ecological, temporally-extended, interpersonal and narrative components set within a multi-layered model of mind.

We consider the interpretation of emotions and similiar phenomena as support for survival and coping in the world. Grounded in the first-person experience of an emotional agent, certain such emotions, drives or experiences are self-oriented (homeostasis, intake, outflow: hunger, pain, irritation), while others suggest a generalized or specific recognition of other agents or objects (curiosity, fear; or hatred, envy, yearning, greed). Other, more complex emotions are involved in relations to a second person (sympathy) or social regulation (shame, guilt, feelings of loyalty) or affective episodic structure (hope, regret). Considering complex emotions in relation to other ‘persons’ yields insight into the roles and possible design of various emotional phenomena in behavioral regulation in biological, software, and social contexts. Affective coloring of episodic memories of sequences of actions and experiences may suggest a mechanism for the grounding of behavioral adaptation. We explore channels of meaning for agents in interaction games as these relate to emotions, the temporal dynamics of affect in relation to behavior, remembering, and learning; and we outline how affective coloring of episodic memories might provide a mechanism for emergent spatial and social navigation, as well as considering the role of the temporal horizon in behavior selection.

The past two decades have witnessed a resurgence of interest in emotion research, as well as progress in understanding the circuitry that mediates affective processing in biological agents. Emotion researchers are now recognizing that computational models of emotion provide an important tool for understanding the mechanisms of affective processing. There has also been significant progress in affective computing technologies, including affective virtual agents, social robots and affect-adaptive human-computer interaction in general, including affective gaming and the associated desire to model more affectively realistic and believable agents and robots. This chapter describes a generic methodology for modeling emotions and their effects on cognitive processing. The methodology is based on the assumption that a broad range of both state and trait influences on cognition can be represented in terms of a set of parameters that control processing within the architecture modules. As such, the methodology is suitable both for exploring the nature of the mechanisms mediating cognition-emotion interaction and for developing more affectively realistic and believable agents and robots. An implementation of this generic methodology in a symbolic cognitive-affective architecture is described, focusing on an example of a research model. The chapter concludes with a discussion of open questions and challenges in affective modeling.

This work proposes a decision-theoretic approach to problems involving interaction between robot systems and human users, with the goal of estimating the human state from observations of its behavior, and taking actions that encourage desired behaviors. The approach is based on the Partially Observable Markov Decision Process (POMDP) framework, which determines an optimal policy (mapping beliefs onto actions) in the presence of uncertainty on the effects of actions and state observations, extended with information rewards (POMDP-IR) to optimize the information-gathering capabilities of the system. The POMDP observations consist of human gestures and spoken sentences, while the actions are split into robot behaviors (such as speaking to the human) and information-reward actions to gain more information about the human state. Under the proposed framework, the robot system is able to actively gain information and react to its belief on the state of the human (expressed as a probability mass function over the discrete state space), effectively encouraging the human to improve his/her behavior, in a socially acceptable manner. Results of applying the method to a real scenario of interaction between a robot and humans are presented, supporting its practical use.

DIARC has been under development for over 15 years. Different from other cognitive architectures like SOAR or ACT-R, DIARC is an intrinsically component-based distributed architecture scheme that can be instantiated in many different ways. Moreover, DIARC has several distinguishing features, such as affect processing and deep natural language integration, is open-world and multi-agent enabled, and allows for “one-shot instruction-based learning” of new percepts, actions, concepts, rules, and norms. In this chapter, we will present an overview of the DIARC architecture and compare it to classical cognitive architectures. After laying out the theoretical foundations, we specifically focus on the action, vision, and natural language subsystems. We then give two examples of DIARC configurations for “one-shot learning” and “component-sharing”. We also briefly mention different use cases of DIARC, in particular, for autonomous robots in human-robot interaction experiments and for building cognitive models.

Social robots have the potential to problematize many attributes that have previously been considered, in philosophical discourse, to be unique to human beings. Thus, if one construes the explicit programming of robots as constituting specific objectives and the overall design and structure of AI as having aims, in the sense of embedded directives, one might conclude that social robots are motivated to fulfil these objectives, and therefore act intentionally towards fulfilling those goals. The purpose of this paper is to consider the impact of this description of social robotics on traditional notions of intention and meaning-making, and, in particular, to link meaning-making to a social ecology that is being impacted by the presence of social robots. To the extent that intelligent non-human agents are occupying our world alongside us, this paper suggests that there is no benefit in differentiating them from human agents because they are actively changing the context that we share with them, and therefore influencing our meaning-making like any other agent. This is not suggested as some kind of Turing Test, in which we can no longer differentiate between humans and robots, but rather to observe that the argument in which human agency is defined in terms of free will, motivation, and intention can equally be used as a description of the agency of social robots. Furthermore, all of this occurs within a shared context in which the actions of the human impinge upon the non-human, and vice versa, thereby problematising Anscombe’s classic account of intention.

This chapter discusses Social Smart Environments (SSEs) with a large number of believable Embodied Conversational Agents (ECAs) in the context of globalization. It focuses on SSE architecture, rapid prototyping and scalability with respect to size, geography, and administration. SSE is a software environment installed in a physical place representing, for example, a city inhabited by believable ECAs that interact comprehensibly with each other; believable ECAs are software agents that stand for humans from different cultures. To ensure believability, the ECAs maintain various determinants of processing, for instance, emotional, personal and cultural, identified through an analysis of 35 scenarios of intercultural interaction. This chapter shows implementation of these determinants and development of an SSE prototype on the basis of a specification defining interaction between ECAs. In conclusion, this contribution provides insight into future work addressing, for example, innovation in societies simulated by SSEs.

This chapter discusses individual-based models (IBMs) and uses the Overview, Design concepts, and Details (ODD) protocol to describe a predator-prey evolutionary ecosystem IBM called EcoSim. EcoSim is one of the most complex and large-scale IBMs of its kind, allowing hundreds of thousands of intricate individuals to interact and evolve over thousands of time steps. Individuals in EcoSim have a behavioral model represented by a fuzzy cognitive map (FCM). The FCM, described in this chapter, is a cognitive architecture well-suited for individuals in EcoSim due to its efficiency and the complexity of decision-making it allows. Furthermore, it can be encoded as a vector of real numbers, lending itself to being part of the genetic material passed on by individuals during reproduction. This allows for meaningful evolution of their behaviors and natural selection without predefined fitness. EcoSim has been enhanced to increase the breadth and depth of the questions it can answer. New features include: fertilization of primary producers by consumers, predator-prey combat, sexual reproduction, sex-linkage of genes, multiple modes of reproduction, size-based dominance hierarchy, and more. In addition to describing EcoSim in detail, we present data from default EcoSim runs to show potential users the types of data EcoSim generates. Furthermore, we present a brief sensitivity analysis of some variables in EcoSim, and a case study that demonstrates research that can be performed using EcoSim. In the case study, we elucidate some evolutionary and behavioral impacts on animals under two conditions: when primary production is limited, and when energy expenditure is reduced.