Distributed Cognition and eLearning

This is one for the cyber-nerds. It's my final assignment submission for the course "E-Learning and Digital Culture" in the MSc in Digital Education at the University of Edinburgh. You can take the module as a MOOC on Coursera. I really recommend it - it was excellent! Here's the essay below, which can also be found on my blog at http://www.danielgriffin.net. Hope you enjoy!

Distributed Cognition and eLearning - Introduction

For over a decade now, there has been ongoing debate and a gradual "shift in cognitive science toward a view of cognition as a property of systems that are larger than isolated individuals. This extends the reach of cognition to encompass interactions between people as well as interactions with resources in the environment", (Dcog-HCI Lab). This is in marked contrast to so called narrow cognition or internalism, which maintains that the cognitive process is a closed phenomena that exists entirely within a distinct mind. This Theory of Distributed Cognition (Hutchins, 1995) "is a branch of cognitive science that proposes that human knowledge and cognition are not confined to the individual. Instead, it is distributed by placing memories, facts, or knowledge on the objects, individuals, and tools in our environment" (1). That is to say, "certain forms of human cognizing include inextricable tangles of feedback, feed-forward and feed-around loops: loops that promiscuously criss-cross the boundaries of brain, body and world", (Clark, 2008). The theory of distributed cognition has been applied to the analysis of everything from US Navy bridge crew operations, aircraft cockpit and instrumentation design, through to commercial business processes, software development teams and other group based work. In this essay I discuss how this framework can be useful in the design of more effective elearning environments. I also speculate on what distributed cognition might mean within systems containing intelligent subsystems and/or artificially intelligent computer applications.

Background to Distributed Cognition
"Distributed Cognition is a hybrid approach to studying all aspects of cognition, from acognitive, social and organisational perspective", (Rogers, 1997). "The Distributed Cognition ... approach was developed by Ed Hutchins and his colleagues at University California, San Diego in the mid to late 80s as a radically new paradigm for rethinking all domains of cognitive phenomena", (Rogers, 1997). Hutchins continued to develop the ideas resulting in his renowned work, Cognition in the Wild, in which he described the evolution of the theory following a yearlong ethnographic study aboard a US Navy steamship. Although some studies have employed close observation and video of test subjects to collect empirical data, the use of ethnographic research is seen as an important tool in understanding cognition as a socially distributed phenomenon because, it is only through embedded observational techniques and direct experience that one may come close to a practical understanding of the actual processes taking place. During the course of his study, Hutchins observed that the collaborative tasks performed by the crew appeared to be carried out as an distributed process composed of the crew themselves, the ships instrumentation, engineering systems and command decisions. Informed by this observation, Hutchins proposed that cognition should not be considered an internal phenomenon, and stated that "instead of focusing on human activity in terms of processes acting upon representations inside an individual actors heads ... [distributed cognition theory] ... seeks to apply the same cognitive concepts, but this time, to the interactions among a number of human actors and technological devices for a given activity" (Rogers, 1997). As such, the "unit of analysis" within distributed cognitive systems is said to be variable, because it is potentially undergoing constant change with new elements possibly entering or exiting the system at any time during processing.

Distributed cognition theory holds that (2):

Cognitive processes may be distributed across the members of a social group.

Cognitive processes may be distributed in the sense that the operation of the cognitive system involves coordination between internal and external (material or environmental) structure.

Processes may be distributed through time in such a way that the products of earlier events can transform the nature of related events.

(Wikipedia, 2012)

Distributed Cognition and Connectivism

Hutchins theory has heavily influenced the ideas of Connectivism, which "sees learning as the process of creating connections and developing a network" (3). Specifically, Distributed Cognition informs the Connectivist ideas that learning occurs through the building of connections, much like connecting nodes in a network; knowledge may exist outside of the human mind, and that nurturing and maintaining the network connections which define interactions between the internal and external components of the system (nodes), is essential for learning to take place. Nodes within the network may take many forms, from individual persons and their thoughts, feelings and social interactions as well as the tools they use.

Elements and characteristics of a network include

Content (data or information)

Interaction (tentative connection forming)

Static nodes (stable knowledge structure)

Dynamic nodes (continually changing based on new information and data)

Self-updating nodes (nodes which are tightly linked to their original information source, resulting in a high level of currency

Emotive elements (emotions that influence the prospect of connection and hub formations).

(Seimens, 2005)

Although connectivism has been criticized as being simply a "pedagogical view" (Verhagen, 2006) rather than an actual learning theory, it "continues to play an important role in the development and emergence of new pedagogues, where control is shifting from the tutor to an increasingly more autonomous learner", (Kop and Hill, 2008).

Learning environment design informed by Distributed Cognition Theory

Stahl, Koschmann & Suthers (2006) offer a socio-cultural perspective on cognition, seeing it as being something that is inseparable from the context in which it happens, however others have argued that this is an oversimplification of the actual processes taking place. "Events are also occurring at a group level that cannot be explained by investigation at the individual level” (Claro, 2011). It would appear that cognition takes place simultaneously on two levels, both at the individual level, and the group level. This idea strongly supports Hutchins views on the socially distributed nature of cognition. Groups and organizations build a common cognitive process that supports not just the group but also the individual members. Harris tells us that "a distributed cognition analysis of memory in cooperative organizations reveals several aspects of organizational memory that seem useful from the standpoint of HCI design", (Harris, undated). Such organizational interactions can now be readily modeled in software, allowing their benefits to be reproduced in training applications.

Human training within adaptive computer assisted learning environments allows for the possibility of keeping the learner in a constant state of "flow", a state of awareness that occurs when someone is neither bored nor over stimulated, but experiencing a perfect balance between the task at hand and their level of ability. By carefully monitoring a learners progress and understanding of a concept, embedded intelligent systems working within our virtual learning environments can adapt any learning material to fit that particular students' needs in real time.

We are facing a time when the currency of information within certain domains is measured in increasingly shorter time periods. What was established fact yesterday may not be such tomorrow and therefore "pattern recognition is the new critical skill" (Siemens, 2006). Though the use of active tools embedded within learning environments, we can offer learners the possibility of seeing through many different windows onto the same information, or seeing many different information sets from similar perspectives. But more importantly we can offer them the understanding of how to make connections and identify relevant data - software has the potential to teach pattern recognition rather than merely supporting rote learning. Software learning environments can be highly adaptable and capable of customization for individual learning styles, as well as being simple to update and reconfigure when further challenge is required by the learner. The idea of active tools supporting a user is not entirely new to learning theory, and there are many examples of similar components within computer systems. Continual advances in AI and machine learning have allowed software systems to become at least as reliable as humans for certain types of tasks, and orders of magnitude more efficient and/or effective for many others.

The following list are examples of the types of software applications that can support a user when working with a computer system, and which might easily be adapted for use in the design of computer enhanced and computer assisted learning environments.

Learning Support Tools

As I write this text, my software word processor is automatically saving my work at regular intervals. And while I am still conscious of the need to save my progress, I have chosen to allow the software to manage this menial task for me. Thus one could claim that I am not merely using the software programme, but that the programme and I are actually working together. It has become a part of my cognitive model and something which I have come to rely on and expect it to perform certain functions for me. The actual process that it uses is unimportant to me, as long as the end result is a saved file that I can rely on. Learning support tools may be as varied as the specific tasks they aim to assist with. Everything from search engines and document management systems, through to reminders and alarm clocks can be considered within this category of tools. By embedding intelligent tools into our learning environments and our learning management systems, we extend their power and allow for greater customization than ever before. And while the idea of assisting learners with software is not new (for a current example, pick nearly any extension module to the Moodle LMS(4)), we are reaching a point where our tools are waking up and watching us. They can provide remediation, branching and increased emphasis where learners are observed to be experiencing difficulties. They may suggest further reading based on web 2.0 style up-voting or even be configured to insist that learners redo certain sections of course material if their understanding is incomplete. This need not be based merely on student test scores, but on an actual personalized model of the students understanding of a subject as observed by their progress through the course material.

Expert Systems

Expert systems are software applications designed to solve complex problems that would require extensive human knowledge in a particular domain. They do this by replicating the type of reasoning that humans use to arrive at solutions, inferring from one step to the next through the use of IF-THEN rules and facts, however they add to this vast datasets of relevant information and are very often connected to large online databases. Expert systems can grow in complexity with the addition of new rules and more data, and of course can easily work on many similar problems simultaneously.

Wizards

Wizards are software tools designed to lead a user through a certain set of steps. The typical example is that of a set-up or installer program, designed to help a user activate and configure a larger software application. Software wizards use a set of dialogs arranged in a logical tree structure such that the user can be guided towards their goal without needing to understanding any of the low level details of the steps that they are preforming. In essence, software wizards provide an abstraction layer and a "friendly" interface to complex or difficult systems. They allow the user to ignore much of the complexity of a given task and to simply specify what the end result should be rather than working through the discrete low level steps required to achieve it. In this sense they act as a type of cognitive prosthesis, extending the power of the user far beyond the limits of their natural abilities. When using wizards one gets the feeling that you are relinquishing some amount of control over the micromanagement of the system in order to gain greater control over the system as a whole. Thus wizards become trusted tools that the user can expect to perform a task for them.

Software Agents

Agents offer the user a trusted interface to an application suite or a set of services. Generally agents are presented to the user as a graphical character or avatar, or perhaps a text-only chatbot interface with which they may interact. It has been claimed by Clifford Nass and Byron Reeves (1995) of Stanford University, that such interfaces "attempt to exploit [the] human desire for socially competent technology ... by incorporating many features of 'natural environments and social relations.'" It is now well understood (Chomsky, 2006) that the processes of dialog and social interaction are important, if not crucial components in the development of human intellect, reasoning and cognition. Software agents leverage this knowledge to provide a more natural interface between the user and the machine. Traditional examples of software agents include Microsoft's Bob and Clipit characters, as well as a growing number of web based agents that can be employed by software developers to interface with online systems (for example Automated Online Assistants - eg Paypal and Coke websites). Aggregations of various software services, database services and network enabled systems, plus any of a growing number of other online commercial support tools can easily be accessed via artificially intelligent software agents. Interaction with computer systems (and of course learning systems) via software agents allows the user to work in a fashion that they are accustomed to. "Cultural practices are a key component of human cognition." (Hutchins, 2011) and working in a way that simulates these interactions allows learners to grasp concepts far more easily than simply reading text or passively watching video, listening to audio, etc.

Daemons

Similar to the concept of an agent but acting in an entirely autonomous mode, Daemons are software programs that perform tasks either for the benefit of the user or for the entire system. Most users will be able to ignore any daemons running within their computers, content in the knowledge that the tasks will be carried out for them without requiring any user input.

Cognition distributed across human / computer hybrid systems

If our cognition exists as a field effect rather than something seated entirely within the brain, how might our cognitive processing power be enhanced when the objects comprising this field are themselves highly capable and able to automate complex tasks on our behalf? "Tools today serve a purpose that is largely based on the 'old' model of library catalogue and encyclopedia. As categorization (and finding) models, they serve a purpose when we have a one-dimensional relationship to knowledge (namely that we understand we need it and, in the process, seek to acquire it). What happens when software/technology does this for us? What happens when the knowledge we require is presented to us without having to consciously seek it (artificial intelligence)?", (Siemens, 2006). If the cognitive props and that we rely on are actually actors preforming independent roles and providing services, might we not be on the verge of a potentially profound jump in the level of our cognitive powers? "The importance of this new perspective is profound. If our minds themselves can include aspects of our social and physical environments, then the kinds of social and physical environments we create can reconfigure our minds and our capacity for thought and reason." (Oxford University Press, 2011). Putting aside fashionable arguments as to whether or not we are already posthuman (5), such an evolution as this would undoubtedly change us in ways we cannot yet even imagine.

The evidence to support these ideas is becoming increasingly more visible in the digital age. At a time when network enabled software tools are now readily available, we are beginning to see the fruits of socially distributed cognition taking place in real time and "in the wild". A prime example of this is the aggregation of crowdsourced tagging into types of group created taxonomies know as folksonomies; thus allowing for socially produced tag clouds, web indexing and meta data creation. When knowledge is produced in this fashion it seems almost as if it is an emergent property of the network, a synergistic boost in output created through the aggregated production and symphony of a million interconnected minds and the software tools they wield.

It is very tempting to draw parallels between the processes at work in socially distributed cognition and those of distributed computing systems. In many ways the models used in distributed computing represent an analogue of the natural processes that take place when groups of individuals work collaboratively on problem solving and information processing. Problems are broken down into manageable sub segments and solved independently before being used to determine an overall result. The final output is therefore produced by a type of gestalt composed of many algorithms for distributed computers, or many minds and their tools in a (socially) distributed cognitive system. It is only when we start to consider what the combination of both types of system can produce; organic and digital working in harmony, that we begin to see the enormous power that the network is capable of. Laborious tasks that humans find difficult can be "farmed out" to the computer, while human input can help to resolve issues that are either too difficult or computationally expensive for computers to effectively process. A great example of this type of system are human-based computation games, also known as games with a purpose (6). Such systems use humans as a type of integral system component, employed to produce a desired end result. For example, one long standing problem in computer science which is very difficult to solve with mere algorithms and processing power, is that of identifying objects in images. Various techniques such as measuring colour changes across an image have been employed without much success in the past; because this is a problem of semiotics and the interpretation of symbols rather than something that can be solved by referencing a library of known patterns. An innovative technique to solve this difficult problem was originally proposed by Dr. Luis von Ahn as “The ESP Game” and subsequently licensed by Google Inc in 2006. It involves posing the problem as a game to be played by two human players. Each player is presented with the same image and must pick a single word to describe it. If both players select the same word then this can be considered a good label for the image. When deployed as an online game and opened to the web, the result is a crowdsourced, people powered labeling system for online image libraries. Usually when we speak about human computer interaction, we are talking about what Angus et al have referred to as "power relations" (Angus et al, 2001) with the human clearly playing a controlling role. However in such a system as this, humans and computers might be considered to be working symbiotically and reciprocally, rather than one as the mere tool of the other.

Conclusion

Given the plastic nature of the human mind combined with the every changing electronic environments in which it must interact, perhaps one theory alone will never be adequate to fully describe the complexities of cognition. However by paying careful attention to the dynamics of the cognitive process we may at least produce an effective model. It is my view that the framework proposed by Distributed Cognition provides sufficient flexibility to model this highly dynamic process. "Distributed cognition could help build models of exactly what is going on in the online world. The models that are produced could then be used to help courseware designers and researchers develop systems that reflect what users expect and need." (Jugo, Braidwood, Long, & Stringer, 2011).

However this approach is not without its problems or limitations. Distributed Cognitive Analysis relies heavily of the use of ethnographic type studies of dynamic processes, thus requiring considerable effort and forethought for each bespoke application. "It is not a methodology that one can readily pick off the shelf and apply to a design problem" (Rogers 1997). I suspect that ultimately, "what is needed is not a new stand-alone theory for the digital age, but a model that integrates the different theories to guide the design of online learning materials" (Anderson, 2008). But if we can be sure of anything it is this; that the more we live our lives online and in a constantly connected state, and the more we rely on remote services to inform our thought processes, then the further we extend our minds into the network. Our environments and interactions shape us more than we realise and we must be mindful of the changes they can produce in us. Distributed Cognition Theory offers an insight into how these changes are taking place and how we might best learn and benefit from them.

References

Angus, Tim; Cook, Ian; Evens, James (2001) A Manifesto for Cyborg Pedagogy? University of Birmingham, UK

Anderson (2008) Foundations of Educational Theory for Online Learning, Mohamed Ally. In The Theory and Practice of Online Learning, Terry Anderson, Ed., May 2008

Clark, Andy (2008) Supersizing the Mind. Oxford University Press

Chomsky, Noam (2006) - Language and Mind. Cambridge University Press

Dcog-HCI Lab. Online, retrieved 01.01.12 - http://hci.ucsd.edu/

Gureckis & Goldstone, 2006; Thompson & Fine, 1999, quoted in Claro, Jennifer. (2011) Online, retrieved 01.01.2012. http://jenniferclaro.wordpress.com/2011/06/13/cognitive-and-socio-cultural-perspectives-on-learning-can-the-two-be-reconciled/

Harris, Steven (Undated) Online, retrieved 01.01.2012 http://mcs.open.ac.uk/yr258/dist_cog/

Hutchins, E. (1995) Cognition in the wild. MIT Press: Cambridge, MA.

Hutchins, E. (2011) Online, retrieved 01.01.12. http://hci.ucsd.edu/hutchins/

Jugo , Gordana; Braidwood, Jaki; Long, Jennifer & Stringer, John (2011). Social Approaches to Learning. Oonline, retrieved 02.01.2012 - https://sites.google.com/site/socialapproachestolearning/distributed-cognition/distributed-cogtheory-with-practice/

Kop, Rita and Hill, Adrian (2008). Connectivism: Learning theory of the future or vestige of the past? The International Review of Research in Open and Distance Learning, Vol 9, No 3

Nass , Clifford and Reeves, Byron (1995) Stanford University. Online , retrieved 01.01.2012 http://www.stanford.edu/dept/news/pr/95/950106Arc5423.html

Oxford University Press website review of Supersizing the Mind. Online, retrieved 30.12.2011 http://www.oup.com/us/catalog/general/subject/Philosophy/Mind/?view=usa&ci=9780195333213

Rogers, Yvonne (1997) A Brief Introduction to Distributed Cognition

Seimens, George (2005) Online, retrieved 01.01.12.http://www.astd.org/LC/2005/1105_seimens.htm

Siemens, George (2006) Knowing Knowledge http://www.knowingknowledge.com

Stahl, Koschmann & Suthers. (2006) Quoted in Claro, Jennifer. (2011) Online, retrieved 01.01.2012. http://jenniferclaro.wordpress.com/2011/06/13/cognitive-and-socio-cultural-perspectives-on-learning-can-the-two-be-reconciled/

Verhagen, Pløn. (2006) Connectivism: a new learning theory? University of Twente

Bibliography

1 & 2 http://en.wikipedia.org/wiki/Socially_distributed_cognition, retrieved online – 02.01.2012

https://en.wikipedia.org/wiki/Connectivism, retrieved online – 02.01.2012
http://moodle.org

Hayles, Katherine (1999) How we became posthuman:virtual bodies in cybernetics, literature, and informatics

More, Max (1994) On Becoming Posthuman. http://www.maxmore.com/becoming.htm

Pepperell, Robert (1995) The post-human condition. Intellect books

http://en.wikipedia.org/wiki/Human-based_computation_game