Pragmatics & Cognition - Volume 14, Issue 2, 2006

Some of the papers in this Special Issue distribute cognition between what is going on inside individual cognizers’ heads and their outside worlds; others distribute cognition among different individual cognizers. Turing’s criterion for cognition was for individual, autonomous input/output capacity. It is not clear that distributed cognition could pass the Turing Test.

As is often the case when scientific or engineering fields emerge, new concepts are forged or old ones are adapted. When this happens, various arguments rage over what ultimately turns out to be conceptual misunderstandings. At that critical time, there is a need for an explicit reflection on the meaning of the concepts that define the field. In this position paper, we aim to provide a reasoned framework in which to think about various issues in the field of distributed cognition. We argue that both relevant concepts, distribution and cognition, must be understood as continuous. As it is used in the context of distributed cognition, the concept of distribution is essentially fuzzy, and we will link it to the notion of emergence of system-level properties. The concept of cognition must also be seen as fuzzy, but for a different reason: due to its origin as an anthropocentric concept, no one has a clear handle on its meaning in a distributed setting. As the proposed framework forms a space, we then explore its geography and (re)visit famous landmarks.

Synthesizing the domains of investigation highlighted in current research in distributed cognition and related fields, this paper offers an initial taxonomy of the overlapping types of resources which typically contribute to distributed or extended cognitive systems. It then outlines a number of key dimensions on which to analyse both the resulting integrated systems and the components which coalesce into more or less tightly coupled interaction over the course of their formation and renegotiation.

Humans are closely coupled with their environments. They rely on being ‘embedded’ to help coordinate the use of their internal cognitive resources with external tools and resources. Consequently, everyday cognition, even cognition in the absence of others, may be viewed as partially distributed. As cognitive scientists our job is to discover and explain the principles governing this distribution: principles of coordination, externalization, and interaction. As designers our job is to use these principles, especially if they can be converted to metrics, in order to invent and evaluate candidate designs. After discussing a few principles of interaction and embedding I discuss the usefulness of a range of metrics derived from economics, computational complexity, and psychology.

A strong case can be made that the cognitive system is designed for guiding action, not, for example, symbol manipulation. I review empirical work demonstrating the link between action and cognition with special attention to the processes of language comprehension. Next, I sketch an embodied cognition framework for integrating work on language understanding with a more general approach to cognition and action. This general approach considers contributions to action of bodily states, emotions, social and cultural processes, and learning within a framework that generates a dynamic system. This framework is used to consider the notion of distributed cognition and the prospects that technology might induce substantial changes in cognition. My assessment is that such changes are unlikely.

The double function of language, as a social/communicative means, and as an individual/cognitive capability, derives from its fundamental property that allows us to internally re-represent the world we live in. This is possible through the mechanism of symbol grounding, i.e., the ability to associate entities and states in the external and internal world with internal categorical representations. The symbol grounding mechanism, as language, has both an individual and a social component. The individual component, called the “Physical Symbol Grounding”, refers to the ability of each individual to create an intrinsic link between world entities and internal categorical representations. The social component, called “Social Symbol Grounding”, refers to the collective negotiation for the selection of shared symbols (words) and their grounded meanings. The paper discusses these two aspects of symbol grounding in relation to distributed cognition, using examples from cognitive modeling research on grounded agents and robots.

The paper discusses recent developments in web technologies based on collaborative tagging. This approach is seen as a tremendously powerful way to coordinate the ontologies and views of a large number of individuals, thus constituting the most successful tool for distributed cognition so far.

Is cognition an exclusive property of the individual or can groups have a mind of their own? We explore this question from the perspective of complex adaptive systems. One of the principal insights from this line of work is that rules that govern behavior at one level of analysis (the individual) can cause qualitatively different behavior at higher levels (the group). We review a number of behavioral studies from our lab that demonstrate how groups of people interacting in real-time can self-organize into adaptive, problem-solving group structures. A number of principles are derived concerning the critical features of such “distributed” information processing systems. We suggest that while cognitive science has traditionally focused on the individual, cognitive processes may manifest at many levels including the emergent group-level behavior that results from the interaction of multiple agents and their environment.

If distributed cognition is to become a general analytic frame, it needs to handle more aspects of cognition than just highly efficient problem solving. It should also handle learning. We identify four classes of distributed learning: induction, repurposing, symbiotic tuning, and mutual adaptation. The four classes of distributed learning fit into a two-dimensional space defined by the stability and adaptability of individuals and their environments. In all four classes of learning, people and their environments are highly interdependent during initial learning. At the same time, we present evidence indicating that certain types of interdependence in early learning, most notably mutual adaptation, can help prepare people to be less dependent on their immediate environment and more adaptive when they confront new environments. We also describe and test examples of learning technologies that implement mutual adaptation.

This article describes a representation-based framework of distributed cognition. This framework considers distributed cognition as a cognitive system whose structures and processes are distributed between internal and external representations, across a group of individuals, and across space and time. The major issue for distributed research, under this framework, are the distribution, transformation, and propagation of information across the components of the distributed cognitive system and how they affect the performance of the system as a whole. To demonstrate the value of this representation-based approach, the framework was used to describe and explain an important, challenging, and controversial issue — the concept of affordance.

Theories on categorization have led to numerous technical innovations. Starting with artificial intelligence and neural models, scientists have leveraged psychological theories to drive forward innovative technology. More recently, software companies and Internet firms have implemented high technology software developed from cognitive theory. One class of systems rooted in the philosophical tradition stresses the importance of explanation and function. Another focuses on feature similarity and rule-based reasoning. Both approaches have had modest success and solve fundamental problems, but neither has achieved the higher-levels of categorization found in humans. The current paper critically analyzes the various theories of categorization and their collective impact on recent technology innovations.

Crime scene investigation is a form of Distributed Cognition. The principal concept we explore in this paper is that of ‘resource for action’. It is proposed that crime scene investigation employs four primary resources-for-action: (a.) the environment, or scene itself, which affords particular forms of search and object retrieval; (b.) the retrieved objects, which afford translation into evidence; (c.) the procedures that guide investigation, which both constrain the search activity and also provide opportunity for additional activity; (d.) the narratives that different agents within the system produce to develop explanatory models and formal accounts of the crime. For each aspect of distributed cognition, we consider developments in technology that could support activity.

I analyse the impact of search engines on our cognitive and epistemic practices. For that purpose, I describe the processes of assessment of documents on the Web as relying on distributed cognition. Search engines together with Web users, are distributed assessment systems whose task is to enable efficient allocation of cognitive resources of those who use search engines. Specifying the cognitive function of search engines within these distributed assessment systems allows interpreting anew the changes that have been caused by search engine technologies. I describe search engines as implementing reputation systems and point out the similarities with other reputation systems. I thus call attention to the continuity in the distributed cognitive processes that determine the allocation of cognitive resources for information gathering from others.

This paper outlines the background development of “intelligent” technologies such as speech recognition. Despite significant progress in the development of these technologies, they still fall short in many areas, and rapid advances in areas such as dictation are actually stalled. In this paper we have proposed semi-automatic solutions — smart integration of human and intelligent efforts. One such technique involves improvement to the speech recognition editing interface, thereby reducing the perception of errors to the viewer. Other techniques that are described in the paper are batch enrollment, which allows the user to reduce the amount of time required for enrollment, and content spotting, which can be used for applications that have repeated content flow, such as movies or museum tours. The paper also suggests a general concept of distributive training of speech recognition systems that is based on data collection across a network.

The present article examines the potential effects on the translation process of working interactively with a translation memory (TM) system, a tool for storing and sharing previous translations. A TM system automatically divides the source text into sentences presented to the translator one-by-one. Based on observations made in an empirical study of six professional translators and six translation students, it is argued that full sentences do not constitute a central cognitive processing category in translation, and that the sentence-by-sentence presentation inherent in TM systems therefore creates an unnaturally strong focus on the sentence, which affects the very task of translation (as well as the translation product). Particular attention is given to the impact of the use of TM systems on the informants’ revision behaviour and their tendency to change the sentence structure.