Cognitive architectures: Research issues and challenges

In this paper, we examine the motivations for research on cognitive architectures and review some candidates that have been explored in the literature. After this, we consider the capabilities that a cognitive architecture should support, some properties that it should exhibit related to representation, organization, performance, and learning, and some criteria for evaluating such architectures at the systems level. In closing, we discuss some open issues that should drive future research in this important area.     

The role of falsification in the development of cognitive architectures: insights from a lakatosian analysis

It has been suggested that the enterprise of developing mechanistic theories of the human cognitive architecture is flawed because the theories produced are not directly falsifiable. Newell attempted to sidestep this criticism by arguing for a Lakatosian model of scientific progress in which cognitive architectures should be understood as theories that develop over time. However, Newell's own candidate cognitive architecture adhered only loosely to Lakatosian principles. This paper reconsiders the role of falsification and the potential utility of Lakatosian principles in the development of cognitive architectures. It is argued that a lack of direct falsifiability need not undermine the scientific development of a cognitive architecture if broadly Lakatosian principles are adopted. Moreover, it is demonstrated that the Lakatosian concepts of positive and negative heuristics for theory development and of general heuristic power offer methods for guiding the development of an architecture and for evaluating the contribution and potential of an architecture's research program.     

Execution architectures for program algebra

We investigate the notion of an execution architecture in the setting of the program algebra PGA, and distinguish two sorts of these: analytic architectures, designed for the purpose of explanation and provided with a process-algebraic, compositional semantics, and synthetic architectures, focusing on how a program may be a physical part of an execution architecture. Then we discuss in detail the Turing machine, a well-known example of an analytic architecture. The logical core of the halting problem—the inability to forecast termination behavior of programs—leads us to a few approaches and examples on related issues: forecasters and rational agents. In particular, we consider architectures suitable to run a Newcomb Paradox system and the Prisoner's Dilemma.     

Integrated versus modular theories of number skills and acalculia

This paper contrasts two views of the cognitive architecture underlying numerical skills and acalculia. According to the abstract-modular theory (e.g., McCloskey, Caramazza, & Basili, 1985), number processing is comprised of independent comprehension, calculation, and production subsystems that communicate via a single type of abstract quantity code. The alternative, specific-integrated theory (e.g., Campbell & Clark, 1988), proposes that visuospatial, verbal, and other modality-specific number codes are associatively connected as an encoding complex and that different facets of number processing generally involve common, rather than independent, processes. The hypothesis of specific number codes is supported by conceptual inadequacies of abstract codes, format-specific phenomena in calculation, the diversity of acalculias and individual differences in number processing, lateralization issues, and the role of format-specific codes in working memory. The integrated, associative view of number processing is supported by the dependence of modular views on abstract codes and other conceptual inadequacies, evidence for integrated associative networks in calculation tasks, acalculia phenomena, shortcomings in modular architectures for number-processing dissociations, close ties between semantic and verbal aspects of numbers, and continuities between number and nonnumber processing. These numerous logical and empirical considerations challenge the abstract-modular theory and support the encoding-complex view that number processing is effected by integrated associative networks of modality-specific number codes.     

EARSHOT: A Minimal Neural Network Model of Incremental Human Speech Recognition

Despite the lack of invariance problem (the many-to-many mapping between acoustics and percepts), human listeners experience phonetic constancy and typically perceive what a speaker intends. Most models of human speech recognition (HSR) have side-stepped this problem, working with abstract, idealized inputs and deferring the challenge of working with real speech. In contrast, carefully engineered deep learning networks allow robust, real-world automatic speech recognition (ASR). However, the complexities of deep learning architectures and training regimens make it difficult to use them to provide direct insights into mechanisms that may support HSR. In this brief article, we report preliminary results from a two-layer network that borrows one element from ASR, long short-term memory nodes, which provide dynamic memory for a range of temporal spans. This allows the model to learn to map real speech from multiple talkers to semantic targets with high accuracy, with human-like timecourse of lexical access and phonological competition. Internal representations emerge that resemble phonetically organized responses in human superior temporal gyrus, suggesting that the model develops a distributed phonological code despite no explicit training on phonetic or phonemic targets. The ability to work with real speech is a major advance for cognitive models of HSR.     

Neural blackboard architectures of combinatorial structures in cognition

Human cognition is unique in the way in which it relies on combinatorial (or compositional) structures. Language provides ample evidence for the existence of combinatorial structures, but they can also be found in visual cognition. To understand the neural basis of human cognition, it is therefore essential to understand how combinatorial structures can be instantiated in neural terms. In his recent book on the foundations of language, Jackendoff described four fundamental problems for a neural instantiation of combinatorial structures: the massiveness of the binding problem, the problem of 2, the problem of variables, and the transformation of combinatorial structures from working memory to long-term memory. This paper aims to show that these problems can be solved by means of neural "blackboard" architectures. For this purpose, a neural blackboard architecture for sentence structure is presented. In this architecture, neural structures that encode for words are temporarily bound in a manner that preserves the structure of the sentence. It is shown that the architecture solves the four problems presented by Jackendoff. The ability of the architecture to instantiate sentence structures is illustrated with examples of sentence complexity observed in human language performance. Similarities exist between the architecture for sentence structure and blackboard architectures for combinatorial structures in visual cognition, derived from the structure of the visual cortex. These architectures are briefly discussed, together with an example of a combinatorial structure in which the blackboard architectures for language and vision are combined. In this way, the architecture for language is grounded in perception. Perspectives and potential developments of the architectures are discussed.     

An architecture for ethical robots inspired by the simulation theory of cognition

The expanding ability of robots to take unsupervised decisions renders it imperative that mechanisms are in place to guarantee the safety of their behaviour. Moreover, intelligent autonomous robots should be more than safe; arguably they should also be explicitly ethical. In this paper, we put forward a method for implementing ethical behaviour in robots inspired by the simulation theory of cognition. In contrast to existing frameworks for robot ethics, our approach does not rely on the verification of logic statements. Rather, it utilises internal simulations which allow the robot to simulate actions and predict their consequences. Therefore, our method is a form of robotic imagery. To demonstrate the proposed architecture, we implement a version of this architecture on a humanoid NAO robot so that it behaves according to Asimov’s laws of robotics. In a series of four experiments, using a second NAO robot as a proxy for the human, we demonstrate that the Ethical Layer enables the robot to prevent the human from coming to harm in simple test scenarios.     

Skill transfer through goal-driven representation mapping

In this paper, we present an approach to transfer that involves analogical mapping of symbols across different domains. We relate this mechanism to Icarus, a theory of the human cognitive architecture. Our system can transfer skills across domains hypothesizing maps between representations, improving performance in novel domains. Unlike previous approaches to analogical transfer, our method uses an explanatory analysis that compares how well a new domain theory explains previous solutions under different mapping hypotheses. We present experimental evidence that the new mechanism improves transfer over Icarus’ basic learning processes. Moreover, we argue that the same features which distinguish Icarus from other architectures support representation mapping in a natural way and operate synergistically with it. These features enable our analogy system to translate a map among concepts into a map between skills, and to support transfer even if two domains are only partially analogous. We also discuss our system’s relation to other work on analogy and outline directions for future research.     

Assessing technology acceptance for skills development and real-world decision-making in the context of train driving

Advances in technology have improved operator performance and efficiency in transport but it is not uncommon for end users to resist technology in spite of its benefits. Operators may resist technology from genuine and legitimate concerns though it is often seen as unjustified. While beneficial, such resistance can have detrimental effects on operations and safety, and can result in the withdrawal of a technology. Theories relating to technology acceptance include elements such as perceptions about the purpose and use of the technology, personal impact, individual characteristics, peer influence, perceived equity, and organizational factors. Although considerable research into technology acceptance and resistance has been conducted in other domains, very little has been conducted in transportation. Findings from two Australian studies are reported which examined train driver attitudes to two state-of-the-art technologies aimed at enhancing skills development and real-world decision-making. The technologies were implemented in the form of in-vehicle information support and simulated learning. Analysis of interviews defined three overarching themes relating to technology resistance: task dynamics related to ways of working and safety; redundancy regarding the utility of the technology and the impact on job security; and personal impact with respect to effects on status and the drivers’ capacity to learn new skills. It is argued that domain-specific characteristics must be considered when designing and implementing new technologies to ensure that benefits of technologies are optimised. It is also argued that resistance should be seen as a positive element of the design and implementation process. This paper has high relevance for transport researchers, and practical application for rail organisations and policy makers.     

Systematicity redux

One of the main challenges that Jerry Fodor and Zenon Pylyshyn (Cognition 28:3–71, 1988) posed for any connectionist theory of cognitive architecture is to explain the systematicity of thought without implementing a Language of Thought (LOT) architecture. The systematicity challenge presents a dilemma: if connectionism cannot explain the systematicity of thought, then it fails to offer an adequate theory of cognitive architecture; and if it explains the systematicity of thought by implementing a LOT architecture, then it fails to offer an alternative to the LOT hypothesis. Given that thought is systematic, connectionism can offer an adequate alternative to the LOT hypothesis only if it can meet the challenge. Although some critics tried to meet the challenge, others argued that it need not be met since thought is not in fact systematic; and some claimed not to even understand the claim that thought is systematic. I do not here examine attempts to answer the challenge. Instead, I defend the challenge itself by explicating the notion of systematicity in a way that I hope makes clear that thought is indeed systematic, and so that to offer an adequate alternative to the LOT hypothesis, connectionism must meet the challenge.     

Phonological abstraction in processing lexical-tone variation: evidence from a learning paradigm

There is a growing consensus that the mental lexicon contains both abstract and word-specific acoustic information. To investigate their relative importance for word recognition, we tested to what extent perceptual learning is word specific or generalizable to other words. In an exposure phase, participants were divided into two groups; each group was semantically biased to interpret an ambiguous Mandarin tone contour as either tone1 or tone2. In a subsequent test phase, the perception of ambiguous contours was dependent on the exposure phase: Participants who heard ambiguous contours as tone1 during exposure were more likely to perceive ambiguous contours as tone1 than participants who heard ambiguous contours as tone2 during exposure. This learning effect was only slightly larger for previously encountered than for not previously encountered words. The results speak for an architecture with prelexical analysis of phonological categories to achieve both lexical access and episodic storage of exemplars.     

Learning algorithm for an intelligent decision making system based on multi-agent neurocognitive architectures

The paper presents the formalism of an intelligent decision-making system based on multi-agent neurocognitive architectures, which has an architectural similarity to the human brain. An invariant of the organizational and functional structure of the intellectual decision-making process based on the multi-agent neurocognitive architecture is developed. An algorithm for teaching intelligent decision-making systems based on the self-organization of the invariant of multi-agent neurocognitive architectures is presented. Using this algorithm, an intelligent agent was trained and the architecture of the learning process was built on the basis of an invariant of neurocognitive architecture. Further research is related to training an intelligent agent in more complex behavior and expanding the capabilities of an intelligent decision-making system based on multi-agent neurocognitive architectures.     

Characteristics of Completed Suicides: Implications of Differences Among Methods

Approximately 30% of suicides in New York City are the result of jumping from a height. After describing jumping suicides and jumping sites, we used polychotomous logistic regression to compare the characteristics of suicides by jumping to those of individuals committing suicide by hanging, ingestion, or shooting. Method used was significantly associated with sociodemographics, occupation, and mental health status, even after adjustment for individual access to the means of committing suicide. Our finding of an independent association between personal characteristics and method used provides strong evidence for the hypothesis that controlling access to an agent of suicide can influence overall suicide rates, at least in the short term. Study results support the introduction of preventive programs to control access to commonly employed agents of suicide.     

Mathematics,science and ontology

According to quasi-empiricism, mathematics is very like a branch of natural science. But if mathematics is like a branch of science, and science studies real objects, then mathematics should study real objects. Thus a quasi-empirical account of mathematics must answer the old epistemological question: How is knowledge of abstract objects possible? This paper attempts to show how it is possible. The second section examines the problem as it was posed by Benacerraf in ‘Mathematical Truth’ and the next section presents a way of looking at abstract objects that purports to demythologize them. In particular, it shows how we can have empirical knowledge of various abstract objects and even how we might causally interact with them. Finally, I argue that all objects are abstract objects. Abstract objects should be viewed as the most general class of objects. The arguments derive from Quine. If all objects are abstract, and if we can have knowledge of any objects, then we can have knowledge of abstract objects and the question of mathematical knowledge is solved. A strict adherence to Quine's philosophy leads to a curious combination of the Platonism of Frege with the empiricism of Mill.     

The Debate on the Moral Responsibilities of Online Service Providers

Online service providers (OSPs)—such as AOL, Facebook, Google, Microsoft, and Twitter—significantly shape the informational environment (infosphere) and influence users’ experiences and interactions within it. There is a general agreement on the centrality of OSPs in information societies, but little consensus about what principles should shape their moral responsibilities and practices. In this article, we analyse the main contributions to the debate on the moral responsibilities of OSPs. By endorsing the method of the levels of abstract (LoAs), we first analyse the moral responsibilities of OSPs in the web (LoA_IN). These concern the management of online information, which includes information filtering, Internet censorship, the circulation of harmful content, and the implementation and fostering of human rights (including privacy). We then consider the moral responsibilities ascribed to OSPs on the web (LoA_ON) and focus on the existing legal regulation of access to users’ data. The overall analysis provides an overview of the current state of the debate and highlights two main results. First, topics related to OSPs’ public role—especially their gatekeeping function, their corporate social responsibilities, and their role in implementing and fostering human rights—have acquired increasing relevance in the specialised literature. Second, there is a lack of an ethical framework that can (a) define OSPs’ responsibilities, and (b) provide the fundamental sharable principles necessary to guide OSPs’ conduct within the multicultural and international context in which they operate. This article contributes to the ethical framework necessary to deal with (a) and (b) by endorsing a LoA enabling the definition of the responsibilities of OSPs with respect to the well-being of the infosphere and of the entities inhabiting it (LoA_For).     

A Biologically Plausible Action Selection System for Cognitive Architectures: Implications of Basal Ganglia Anatomy for Learning and Decision‐Making Models

Several attempts have been made previously to provide a biological grounding for cognitive architectures by relating their components to the computations of specific brain circuits. Often, the architecture's action selection system is identified with the basal ganglia. However, this identification overlooks one of the most important features of the basal ganglia—the existence of a direct and an indirect pathway that compete against each other. This characteristic has important consequences in decision‐making tasks, which are brought to light by Parkinson's disease as well as genetic differences in dopamine receptors. This paper shows that a standard model of action selection in a cognitive architecture (ACT‐R) cannot replicate any of these findings, details an alternative solution that reconciles action selection in the architecture with the physiology of the basal ganglia, and extends the domain of application of cognitive architectures. The implication of this solution for other architectures and existing models are discussed.     

A computational unification of cognitive behavior and emotion

Existing models that integrate emotion and cognition generally do not fully specify why cognition needs emotion and conversely why emotion needs cognition. In this paper, we present a unified computational model that combines an abstract cognitive theory of behavior control (PEACTIDM) and a detailed theory of emotion (based on an appraisal theory), integrated in a theory of cognitive architecture (Soar). The theory of cognitive control specifies a set of required computational functions and their abstract inputs and outputs, while the appraisal theory specifies in more detail the nature of these inputs and outputs and an ontology for their representation. We argue that there is a surprising functional symbiosis between these two independently motivated theories that leads to a deeper theoretical integration than has been previously obtained in other computational treatments of cognition and emotion. We use an implemented model in Soar to test the feasibility of the resulting integrated theory, and explore its implications and predictive power in several task domains.     

Workfare: the Subjection of Labour

abstract When viewed as a question of distributive justice the evaluation of workfare typically reflects exclusively on the distribution of income: do the physically capable have a justified claim for state support, or is it fair to demand from those who do work to subsidise this support? Rarely is workfare appraised in terms of how it affects other parties such as employers or other workers, and on the structural effects the pattern of incentives it generates brings about, or as an issue of distributive justice related to a more extensive range of objects of distribution such as access to pleasing jobs. We propose to evaluate workfare by looking at its effects more broadly: (1) we discuss and dismiss on empirical grounds the two most common arguments in favour of workfare, namely that workfare develops self‐reliance among participants and it is more economical in the use of altruism among taxpayers. (2) We suggest that workfare focuses on work as a burden and as a means to income, ignoring other beneficial aspects that are conducive to the worker’s self expression. These other benefits are distributed unevenly between two main groups: the privileged well off and the disadvantaged. (3) we argue that workfare functions as a mechanism to preserve these privileges to one class of workers at the expense of another. Or, to put it the other way, the payment of unconditional welfare to the long term unemployed might be justified as one method of mitigating unjustified privileges.     

Linking somatic and symbolic representation in semantic memory: the dynamic multilevel reactivation framework

Biological plausibility is an essential constraint for any viable model of semantic memory. Yet, we have only the most rudimentary understanding of how the human brain conducts abstract symbolic transformations that underlie word and object meaning. Neuroscience has evolved a sophisticated arsenal of techniques for elucidating the architecture of conceptual representation. Nevertheless, theoretical convergence remains elusive. Here we describe several contrastive approaches to the organization of semantic knowledge, and in turn we offer our own perspective on two recurring questions in semantic memory research: (1) to what extent are conceptual representations mediated by sensorimotor knowledge (i.e., to what degree is semantic memory embodied)? (2) How might an embodied semantic system represent abstract concepts such as modularity, symbol, or proposition? To address these questions, we review the merits of sensorimotor (i.e., embodied) and amodal (i.e., disembodied) semantic theories and address the neurobiological constraints underlying each. We conclude that the shortcomings of both perspectives in their extreme forms necessitate a hybrid middle ground. We accordingly propose the Dynamic Multilevel Reactivation Framework—an integrative model predicated upon flexible interplay between sensorimotor and amodal symbolic representations mediated by multiple cortical hubs. We discuss applications of the dynamic multilevel reactivation framework to abstract and concrete concept representation and describe how a multidimensional conceptual topography based on emotion, sensation, and magnitude can successfully frame a semantic space containing meanings for both abstract and concrete words. The consideration of ‘abstract conceptual features’ does not diminish the role of logical and/or executive processing in activating, manipulating and using information stored in conceptual representations. Rather, it proposes that the materials upon which these processes operate necessarily combine pure sensorimotor information and higher-order cognitive dimensions involved in symbolic representation.