Theory by tyndale


									Computer-supported collaborative learning: a claim for multiple research perspectives in
investigating multiple learner perspectives

Suthers (2005) has shown that CSCL research can be conducted using various epistemologies.
Most notably, he differentiates between a knowledge communication epistemology and a
knowledge construction epistemology1.

Differences in perspective among these epistemologies occur on various levels, viz. the levels of
theory, methodology, and technology.
Researchers advocating a knowledge construction methodology have been very vocal in making
the case for their type of research, and they have formulated many arguments why they think a
“traditional”, knowledge communication epistemology might be inappropriate to address CSCL.
With respect to theory, metholodgy, and technology three general lines of critique can often be
    o Theoretical work should focus on group interactions, and it should try to avoid
         assumptions about mental reprsentations at all cost.
    o Investigations should take a qualitative, hermeneutic approach instead of an experimental
         and/or laboratory and/or quantitative approach.
    o Technologies should be built on constructivist principles instead of principles of
         instructional design. [One should note that this critique does not appear too often
In contrast, researchers advocating a knowledge communication epistemology have rarely
attempted to either defend their position or to show explicit disagreement with views held by
researchers who favor the perspective of knowledge construction. Given this situation, numerous
examples can be found both in research literature and in personal experience where a knowledge
communication epistemology is rejected as being outdated, and the refusal to adopt a knowledge
construction perspective is sometimes interepreted in ways that treat knowledge communication
researchers as not very well-informed on recent CSCL trends.

The main point of this paper is to indicate that while researchers having an affinity to knowledge
communication are aware of strengths and weaknesses of their own methodology, they might
have good reasons not to subscribe to a knowledge construction perspective with regard to CSCL.
While the paper is generally motivated by the desire to have not only peaceful coexistence, but (if
possible) close collaboration among epistemologies, the following list of arguments necessarily
focuses on a critical assessment of the knowledge construction view. It is a direct reply to a very
useful and handy list of arguments presented by Gerry Stahl in preparation for the Tuebingen
Workshop on Group Cognition in October, 2006.

This list of arguments accumulated over a longer time span. It is not held together by a particular
narrative. For purposes of presentation it was ordered according to the three categories of theory,
methodology, and technology.

    The labels associated with these epistemologies can be argued with, but for the sake of brevity I will stick to them.

Scenarios and Metaphors for CSCL
  o All forms of collaborative learning are interesting subjects for scientific inquiry. This
       includes settings where knowledge telling (instead of knowledge building) is prominent,
       or settings where knowledge building and/or knowledge telling fails.
  o It is generally held that groups benefit from a certain diversity among its members.
       Therefore, the notion that collaborative learning is a process leading from unshared
       individual knowledge to shared information serves as a useful working hypothesis. By no
       means this hypothesis precludes that along this trajectory new (external) information or
       new (internal) knowledge is constructed during interaction.

   o CSCL research can greatly benefit from interdisciplinary input. There are numerous
        theoretical approaches that deal with group cognition explicitly or implicitly. Some of
        these are in line with classical CSCL thinking (e.g. shared mental models from
        organizational psychology), while others are in stark contrast (e.g. many social
        psychological accounts).
   o Social psychology tells us a lot about how individuals behave in groups, and how groups
        behave in toto. Social psychology used to be a science of the individual in social settings
        (basically because the individual set of cognitions is the only constant when a person
        participates in various groups). However, in the last 15 years social psychology has
        shown an increased interest in cognition by groups instead of cognition about groups
        (groups as information processors, e.g. Hinsz, Tindale & Vollrath, 1997; Larson &
        Christensen, 1993; Wegner, 1987, 1995)

Unit of Analysis
  o While conversation analysis and other linguistic methods supposedly are based on a
        group level of analysis, the language to describe CSCL processes abounds with examples
        that only make sense with respect to individual entities. E.g., the notion of interaction
        clearly hints that the small group itself or the artifacts generated by the group consist of
        smaller elements. Similarly, the central notion of shared meaning implies and underlines
        that meaning is shared among constitutive elements, i.e. individuals. Given this emphasis,
        it does not appear fruitful to completely abstract away from the obvious, but non-trivial
        fact that utterances are produced by individuals with their idiosyncratic interpretations,
        beliefs, attitudes, skills, and stored experiences – in short, their mental representations.
  o There are several phenomena (e.g. sharedness, conflicts, social dilemmas) that exist only
        on the group level of analysis, and are irreducible to any one individual. However,
        irreducibility does not equal emergence. Collective phenomena in CSCL are “pseudo-
        emergent” rather than emergent, i.e. phenomena on the emergent (group) level can easily
        described using the vocabulary of the sub-emergent (individual) level. To take examples
        from above: an item becomes shared once it is publicly uttered by an individual; a social
        dilemma exists when all individuals have an individual behavioral tendency to maximize
        their profit etc.
  o Many CSCL researcher hold that meaning is relational, and therefore not located in
        individual minds. But this is not necessarily a contradiction. One could argue that
        meaning is an abstract, relational entity that could only be known by an omniscient
       observer. Still it would make much sens to assume that all participants have internal and
       individual assumptions about what the particular meaning is, and these individual
       “instantiations/interpretations” would be the determinants of group interaction. (On a side
       note: the concept of common ground parallels such a line of thinking – common ground is
       abstract, but what counts are individual assumptions about what is part of the common
       ground. In the same vein, meaning is not necessarily irreducible, even if it is relational.

Mental Representations
  o Assuming that mental representations are in any way fixed, is a fundamental
       misconception. E.g., it is firmly established that retrieval from memory is fundamentally
       constructive in nature. The fickle ways in how mental representations can be changed by
       external and internal forces over time is a good reason not to exclude individual cognition
       from the analysis of collaborative learning and group cognition.
  o Verbal expressions, though being dependent on prior utterances, are indicative of mental
       processes. Though short utterances in FTF conversations provide only a glimpse into the
       underlying mental representations, utterances from text-based asynchronous
       communication can best be interpreted as individual attempts to reflect on the current
       state of their mental representations.
  o Focusing on observable interaction is desirable, however, it can never explain the whole
       story. Just as environmental variables are an inextricable context for individual thought,
       individual thought is an inextricable context of collective action. Part of the causal
       explanation of discourse is to analyze why an utterance or gesture was made by a
       particular person. While some of the reasons for making an utterance or gesture are
       dependent on the preceding, observable events, other reasons can only be explained and
       investigated by reference to mental representations. The meaning of a given utterance is a
       function of prior utterances. However, these utterances were interpreted by individuals,
       and the replies made by individuals elaborate on these interpretations by aligning them
       with existing mental representations. Describing and explaining the causal chain of
       collaborative interactions, whether verbal or nonverbal, without recourse to mental
       representations, falls short of distinctive elements of collective activity.


Methodological Scenarios
  o Paraphrasing Roschelle, scientific explanation usually involves a rich set of metaphors
      pointing at a deep-structured situation that abstract from literal features of the world. In
      other words, it is often aimed at uncovering “hidden” mechanisms from observable
      entities. Analyzing transcripts of observable FTF dialogue to uncover uptake events or
      abstract processes of knowledge building is analogous to analyzing tests to uncover the
      abstract concept of knowledge acquisition. In both cases the observable serves to explain
      deep-featured and abstract entities. The difference between these examples is that the
      knowledge test is not a natural by-product of collaborative learning, whereas uptake
      events are a natural by-product.
  o It was claimed that individual and collective analyses of CSCL are mutual and
      complementary accounts that shed a different light on the same phenomenon (cf. the cell
       biological and epidemiological accounts of the flu). However, this analogy is somewhat
       misleading. Whereas cell biology and epidemiology address entirely different aspects of a
       disease, individual and collective analyses of CSCL are aimed at the same aspects, viz.
       the interaction and products created during collaborative learning.
   o   Language use, especially in face-to-face conversations, is highly elliptical, highly
       indexical, and projective. Therefore, video analysis is of great importance to get a detailed
       understanding of the interaction process. However, many if not most CSCL scenarios rely
       on written discourse among people located at different places. It can be assumed that
       ellipsis and indexicality play a much smaller role in these scenarios. Certainly, they are
       still projective. However, projectivity is much more difficult to observe according to
       intersubjectively agreed-upon standards than ellipsis or indexicality.

Causality and Generalizability
  o While inquiries into social sciences are necessarily interpretive in nature, interpretations
       differ in quality. An interpretation that is not intersubjectively shared among researchers,
       or an interpretation that is said to require tacit, indescribable skills of “experts” to be
       accomplished should not be preferred to an interpretation that is intersubjectively shared,
       or an interpretation that can be gained by following precise, and clearly laid-out
       procedural standards.
  o There is causality in single cases, however, this holds only for very simple,
       intersubjectively easily-agreed-upon physical forms of causality (e.g. movement of
       billiard balls). It is highly questionable that the complex causal patterns taking place in
       educational scenarios can be assessed from single case studies. Therefore, the often-
       claimed ability of qualitative methodologies to address causal mechanisms (instead of
       statistical regularities) appears doubtful.
  o Laboratory studies are problematic because they do not take the full context into account.
       This places some limitations on the generalizability of laboratory findings. While case
       studies are more aware of this problem, they do not appear to produce a solution. If every
       case is valid, then even learning taking place in a laboratory would be a legitimate case
       for scientific inquiry. Of course, case studies would be much more adequate if there
       would be clear ideas how to make generalizations from them, an issue that still remains
       very obscure. Though small group interactions are unique and highly situated, it is
       legitimate to attempt to identify principles from a number of such observations. It is
       highly plausible to use experimental methods to warrant some validity to these

Ethnomethology and Conversation Analysis
   o Conversation analysis follows the instruction to observe a phenomenon without recourse
      to theoretical assumptions. This seems paradixical since it is impossible to observe an
      interaction without refererence to pre-existing and pre-interpreted categories. The very
      notion that meaning is relational, that mental representations are negligible, that language
      is always indexical, elliptical and projective, are clear examples of pre-existing theories.
   o The observation that individuals do not know for certain about internal mental states of
      others does not mean that they do not have assumptions about such states. For this reason
      an hermeneutic approach focusing on strictly observable entities is misleading.
      Participants make assumptions about mental states of collaborators all the time. The so-
       called “principal mistake of ethnomethodology”, making assumptions about non-
       observable things, is a good example of our habit to interpret mental states.

Alternative Methodologies
   o Individuals are parts of numerous small groups, and it would not be helpful to think that
       learning gained during an interaction with group A does not transfer to interaction with
       group B. To bridge this gap, individual post-tests are a very useful means in scientific
       investigation, warts and all.
   o To capture knowledge construction, many standard procedures of coding and counting are
       insufficient because they do not take sequential dependencies among utterances into
       account. However, coding and counting per se is not inherently flawed. Qualitative
       studies don’t do counting, but since they cannot be free of theoretical considerations and
       presuppositions, they involve (implicit) coding. What is coded, can also be counted.
       There are means to capture the sequential nature of interactions in quantitative terms.
       Some steps in this direction are provided by the method of sequential analysis although
       this particular method is flawed because it only takes adjacent utterances into account, not
       the larger context.
   o A hybrid methodology using elements of ethnomethodology and psychological
       experimentation could either take the form of analyzing ethnomethodologically across
       different experimental conditions, or to develop quantitative accounts of uptake events.
       While ethnomethodologists would prefer the former hybrid model, experimentalists
       would opt for the latter.
   o Design-based research is a viable and necessary complement to other lines of scientific
       inquiry. Scientific progress should be related to practical applications whenever possible.
       However, most design-based methodologies do not compare methods with alternative
       scenarios (no intervention, other methods), and they do not adhere to standards of
       program evaluation (unbiased, external observers), thus making them difficult to compare
       to alternative designs.

Processes of Knowledge Construction
   o It would be helpful to set standards for what counts as an uptake event and/or knowledge
       construction event. Indexical references and elliptical utterances are linguistic markers
       that can help to get a better grasp of what is going on in an interaction, but they might not
       be sufficient to explain knowledge building. A “good” uptake event for knowledge
       construction activities is an event that cha(lle)nges an existing understanding of a small
   o It is not helpful to say that learning is only acquisition of individual knowledge; likewise,
       it is not helpful to say that learning manifests only during interaction. First and foremost,
       learning is a somewhat lasting change. The problems with regarding learning as
       completely evident in collaborative context are that a) individuals might actively
       “contribute” to knowledge building activities without undergoing any change (e.g.
       students who nod at an explanation without understanding it); b) individuals might get a
       better understanding without leaving any observable trace in transcripts and video
  o CSCL technologies should not necessarily try to replicate unmediated scenarios. A
      “good” CSCL scenario transcends other forms of interaction, and it should strive to
      provide more or better support for groups than would be available in face-to-face
      interactions, if only for dedicated aspects of interaction.
  o Talking about artifacts as carriers of meaning that constrain and afford, i.e. mediate
      actions is a very helpful way to think about technologies for learning. However, one
      should consider to restrict the notion of artifacts to man-made, “designable” entities.
      Artifacts need not be tangible, they can be digital. It is a point for debate whether group
      discussions without any means of recording should be regarded as artifacts. They
      certainly share some characteristics of “real” artifacts, but they certainly differ in others.
      Calling mental representations (cognitive) artifacts is misleading at least. As noted by
      Norman, some mental representations (some heuristics, transferable algorithms) serve as
      tools for thought (cognitive artifacts), but calling all mental representations cognitive
      artifacts does not appear to be plausible and potentially obscures differences between
  o Technology cannot replace teachers, as is exemplified by the intractability of tutoring
      systems. However, regarding technology just as a passive backdrop or medium for human
      interaction, is equally misleading. At its best, technology takes over rudimentary aspects
      of human beings, e.g. by implicitly guiding (instead of prescribing) collective action, or
      by providing helpful background information to a small group, both in terms of content
      and context (e.g. awareness tools)

To top