Knowledge Negotiation in Asynchronous Learning Networks
The negotiation of what is to count as mutually acceptable collaborative knowledge is difficult to conduct when participants cannot interact face-to-face. We review certain related work on negotiation support and develop a concept of “knowledge negotiation” that is appropriate for collaborative learning in ALNs used in school classrooms. This concept is situated within the framework of collaborative knowledge building viewed at the group unit of analysis; it contrasts with negotiation as the reconciliation of multiple personal opinions through voting. We then describe an implementation of support for knowledge negotiation in an ALN that is currently being tested in European schools.
Negotiation is a central phenomenon in cooperative work
and collaborative learning – specifically the negotiation of what is to count
as new shared knowledge. While there has been considerable research on
asynchronous learning networks (ALN) and computer-supported collaborative
learning (CSCL) lately, this has not been accompanied by discussion of computer
software mechanisms to support negotiation within learning contexts.
ALN or CSCL systems are designed to support the building of
shared knowledge, but rarely provide adequate support for establishing and
identifying agreement on achieved knowledge artifacts. Such negotiation is
conceptually different from the forms of negotiation supported in CSCW, GDSS
and other business-oriented systems because in classroom collaborative learning
it is a matter of groups constructing new knowledge interactively, rather than
making decisions based upon pre-defined options and existing opinions of
individuals.
Consideration of computer support for negotiation has arisen
in the past primarily in relation to group decision-support systems (GDSS) for
use in industry. (Connolly, 1997; Kraemer
& Pinsonneault, 1990; Vogel et al.,
1987)
GDSS is a sub-area of computer-supported cooperative work (CSCW). Although CSCW
is a sister field to CSCL, its decision support, knowledge management and
social awareness mechanisms have not yet been adapted for CSCL applications.
This paper provides an example of how one can adapt a CSCW approach to a CSCL
context by re-thinking the nature of the interactions within these differing
contexts. Specifically, we focus on adapting the role of negotiation and
arguing for a concept of “knowledge negotiation.” For a more general discussion
of how we converted a CSCW system to a CSCL system, see (Stahl, 2002b).
Starting in May 2001, the European Union ITCOLE Project
(Information Technology for COllaborative LEarning and knowledge building) (Leinonen et al., 2001) began to develop and test a
new software system for K-12 classroom support named BSCL (Basic Support for
Collaborative Learning). Developers in
BSCL is an adaptation and extension of the well-known and
widely used BSCW system. (Appelt & Klöckner,
1999)
Used by over 200,000 people since 1995 when it was developed at the Institute
for Applied Information Technology – FIT (previously a GMD Institute, now a
Fraunhofer Institute near Bonn, Germany), BSCW provides a system of
autonomously managed Web-based workspaces that can be used by members of a
workgroup to organize and coordinate their work. These workspaces are central
access points for shared documents, including folders for organizing them and a
wealth of functionality for knowledge management.
In BSCL, new components have been added to BSCW to offer
improved support for knowledge building (usually created in threaded
discussions), social awareness (knowing who else is active in the system) and
synchronous interaction (multi-user diagramming with chat). An initial version
of BSCL was successfully field-tested by pedagogic researchers, teachers and
students in
The approach to knowledge negotiation support is integrated
within a set of software components designed for collaborative learning,
including virtual learning spaces, perspectives, community roles, knowledge
building, thinking types and concept maps. Specifically, knowledge negotiation
is implemented to control the publication and transfer of ideas, documents,
drawings and other artifacts or sets of items from a small project group
perspective into the perspective of a larger community of learners in a course.
Knowledge negotiation focuses on evolving a group knowledge
artifact to a mutually acceptable status for publication, rather than reaching
consensus on a pre-existing choice of personal opinions. Asynchronous support
for such negotiation must allow for:
·
the proposal of a set of items for consideration
as a shared knowledge artifact,
·
the discussion of desired modifications to this
artifact,
·
carrying out the actual changes to the items,
·
discussion of remaining misgivings,
·
signaling readiness to accept and publish the
artifact for access by a larger community.
The recognition of a need for support of knowledge
negotiation and the design of BSCL generally is based on classroom experience
with previous ALN systems. Classroom testing of the BSCL base system began in
January 2002. During the summer, the implementation of knowledge negotiation
described here will be released for studies in 50 European classrooms in the
Fall. Preliminary empirical results will be reported at HICSS.
The question we faced as designers of BSCL was how to
support negotiation among students. Collaborative learning in classrooms has
different requirements for sharing knowledge than what is supported by BSCW for
professional teams. For instance, BSCW is used primarily for knowledge
management – the sharing and manipulation of knowledge that already exists
somewhere within the workgroup – while BSCL is intended to support knowledge
building – the collaborative construction of knowledge that is new within the
community.
This paper follows the historical sequence of our approach. We began by considering relevant explorations of negotiation in CSCW (see section 2), particularly those of Herrmann and Wulf that had been used in systems related to BSCW. Then we reflected on the role of negotiation in collaborative learning (sec. 3), based on the major theoretical frameworks for CSCL. From this, we identified various concepts of negotiation associated with alternative possible support mechanisms (sec. 4). We developed a concept of “knowledge negotiation” that seemed most suited for BSCL as an adaptation of BSCW to learning scenarios (sec. 5). This notion may be relevant for many CSCW contexts as well. We implemented support for knowledge negotiation among students in small workgroups (sec. 6), and are now studying negotiation in classrooms using BSCL (sec. 7).
Negotiation is a process by which a group of people who are working together arrive at a group decision. The usual approach to conceptualizing and supporting this process within CSCW was not quite what we wanted for our concept of collaborative knowledge building. We will here review some of the approaches that we critiqued and built from.
Within traditions of computer science (or informatics),
it is common to model negotiation as a voting process. This is not only a
result of the implicit acceptance of rationalist philosophy and of modeling
human communication as information processing, but arises also for pragmatic
implementation reasons:
Rationalism assumes that people have ideas already existing
in their heads (Winograd & Flores,
1986)
– in the form of expressible propositions, mental representations or brain
states – that they can then express verbally as opinions on the basis of which
they may vote on various issues posed to them.
Communication theory derived from the information processing
tradition (Shannon & Weaver,
1949)
implicitly builds on the rationalist model and construes communication as the
transfer of such pre-existing opinions (as data) through (error-prone) media.
Implementation of computer support tends to accept these
models because computers necessarily represent explicit information, such as
propositional representations of explicit opinions. (Stahl, 1993) They can easily respond to
small numbers of clearly pre-defined options, such as yes/no votes.
Thus, when we look for examples of support for negotiation in CSCW, we find that they often reduce negotiation processes to voting processes, assuming that the goal is to collect and respond appropriately to a set of opinions that already exist in the minds of the individual system users. In particular, this is true of GDSS systems that frequently include a component for conducting straw votes. (Connolly, 1997; Kraemer & Pinsonneault, 1990) Straw votes, by definition, are a means of measuring pre-existing personal opinions, with little attempt to influence them or to build group consensus. The goal here is typically to provide support for collecting the opinions of participants about some fixed issue, with the assumption that differences of opinion are based in personal structures of preferences, in differing interests or in limitations of information about the opinions of other participants. (Lim & Benbasat, 1993; McGrath, 1993; Nunamaker Jr. et al., 1991) Thus, GDSS support usually focuses on expressing, collecting and possibly influencing participant opinions, rather than on altering the subject matter under consideration.
Herrmann (Stahl & Herrmann,
1998)
proposed a notion of negotiation that goes significantly beyond the simple
voting model. He and his students developed an approach to computer-supported
negotiation over the years, and have designed and/or prototyped it in a number
of software systems, (Herrmann, 1995;
Herrmann, Wulf, & Hartmann, 1996; Herrmann & Kienle, 2002) including a simulation of
negotiation. (Lepperhoff, 2001) He has reviewed related CSCW
and GDSS research, and has developed a socio-technical model for his approach
to negotiation. His examples involve group decisions for knowledge management,
such as what categories should be used to organize a shared bibliography.
In Herrmann’s approach, someone makes a proposal and the other group members can vote on the proposal. They always have an opportunity to comment on their vote. In addition, they can make a counter-proposal or call for discussion outside of the computer support system. Although this approach goes beyond a simple yes/no voting system with options for counter-proposals and for switching communication media, it is still based on a model of negotiation as voting. This approach serves well to conduct a quick poll to see where agreement does or does not already exist, but cannot well support re-framing or co-construction of knowledge. It recognizes the frequent need for people to engage in more complex processes of interaction to settle a negotiation issue and allows for people to leave the computer support system to do this, but provides little automated support for their consequent decisions to affect the knowledge in the system.
Wulf (Stiemerling & Wulf,
2000; Wulf, 2001; Wulf, Pipek, & Pfeifer, 2001) proposed further extensions
of the voting model, now applied to function activation rather than decisions.
His examples include the right of an individual to access a specific document
created by another member of the group. The empirical cases he cites from
governmental bureaucracies might best be considered examples of moderated,
rather than negotiated, activation. The primary actors do not engage in
negotiation with one another, but agree to have their interactions mediated by
trusted third parties or public procedures, including automated procedures in a
computer support system.
Applied to CSCW systems, the issue is whether a particular user should have access to a specific system function, such as editing a document. Wulf has developed a formal Petri net model of negotiation approaches, but oriented to the question of activation. This paradigm may work for situations with fixed options, such as access to a defined system function, but not in the general situation in which a group is collaborating to produce group knowledge through exploration and inquiry.
Individual learning, as a process of constructing
personal knowledge, takes place within a learner’s personal perspective. (Boland & Tenkasi,
1995; Nygaard & Sørgaard, 1987; Stahl, 1993) Collaborative learning involves
an interaction among personal perspectives contributed by the participants and
a merging of these into a group perspective definitive of the group discourse.
There have been scattered attempts to formulate a conceptualization of
perspectives that would lend itself to computer support. The Phidias system was
an early attempt to display a database of design rationale notes according to
different “contexts”; (McCall et al., 1990) this was subsequently
re-implemented in Hermes, (Stahl, 1993) where shared contents were
displayed within different professional or personal perspectives.
Stahl & Herrmann (Stahl & Herrmann,
1998, 1999)
proposed an approach to integrating Herrmann’s negotiation and Stahl’s
perspective mechanisms within a single software system, WebGuide, that they
designed specifically to explore these mechanisms. The motivation for this was
the following: On the one hand, negotiation takes time, and group members may
want to continue working on a topic while it is under negotiation –
perspectives allows them to continue to work in their own perspective while
contents of a group perspective are being negotiated. On the other hand, within
individual perspectives there is a strong tendency for ideas to diverge (Hewitt & Teplovs,
1999)
– negotiation is required to bring ideas back into consensus and to promote
individual ideas to the status of group knowledge. So it seemed that
integrating perspective and negotiation mechanisms – and conceptualizing
negotiation as the intertwining of multiple personal perspectives to arrive at
a shared perspective – would mutually solve the central problems of these two
mechanisms.
While the perspectives mechanism has by now been extensively implemented in WebGuide, (Stahl, 2001) the corresponding negotiation mechanism is still missing in that system. The lack of an appropriate negotiation mechanism was already reported as a serious limitation of WebGuide at the 1999 CSCL, Group and WebNet conferences (Stahl & Herrmann, 1999). The delay in implementing negotiation support in WebGuide was largely a result of the feeling that the voting model of negotiation did not seem appropriate for CSCL uses of groupware. Recent reflections on the relation of perspectives to knowledge building (Stahl, 2002a) suggest that a different, more dialogical, concept of negotiation is called for.
To appreciate the role of negotiation in CSCL, consider
the centrality of negotiation within each of the different theoretical
frameworks that have historically dominated this field:
Small group process.
This approach maintains a view of learning as transfer of information from
teacher to students, and conducts experiments to demonstrate the increase in
individual learning outcomes through group work in classrooms. A typical
approach would be to divide up topics within a course and assign the topics to
small groups; the small groups would negotiate
agreed upon solutions to their topic; the different groups would then share
their solutions with the larger group, for instance using procedures like
“jig-sawing”. (Brown & Campione,
1994)
Social constructivism.
Knowledge is socially co-constructed (Vygotsky, 1930/1978) before it may
be internalized by children based on what they are capable of understanding.
This social co-construction is a negotiation
process by which shared understanding is reached about a “knowledge object” or
knowledge “artifact”. (Bereiter, 2002; Stahl,
2002a)
Distance education.
Even when peer interaction is possible in distance education, for instance with
threaded discussion in asynchronous learning networks, it is hard to encourage
sustained, in-depth knowledge building; discussions tend to diverge without
some form of negotiation to bring
different people’s ideas back together. (Hewitt & Teplovs,
1999)
Distributed
problem-based learning. Originally developed for medical education, PBL is
built around problem cases, like patients presenting illness symptoms that a
group of about five students and a tutor attempt to diagnose. The group negotiates lists of problem statements,
key evidence, working hypotheses and learning issues. Then the individual
students research relevant medical theories and come back to the group to renegotiate the group understanding. The
tutor plays a key role in guiding the negotiation. (Barrows, 1994)
Distributed cognition.
Knowledge is not simply a matter of an individual’s mental representations, but
is frequently distributed among the abilities of group members and the
artifacts that they use. (Hutchins, 1996) Accordingly, knowledge is
co-constructed by interactions among people and their shared artifacts,
including prominently by means of negotiation
practices that result in establishing a common ground for understanding.
Situated learning.
This approach views learning in terms of changing relations within the
community of practice. (Lave & Wenger,
1991)
Like situated action theory (Suchman, 1987) and ethnomethodology, (Garfinkel, 1967) the situated learning
approach looks at how people skillfully interact socially to co-construct and
interactively negotiate knowledge,
rather than at individuals as possessors of explicit propositional knowledge.
Cultural-historical
activity theory. Learning is viewed as it takes place over extended periods
of time and within its broad cultural and historical contexts. It is even
possible to track “expansive learning” in which multiple groups negotiate changes to the existing social
arrangements. [36] Here, again, socially
shared artifacts play a significant role in providing a focus to
negotiations.
It is possible to conceptualize collaborative learning in different ways, focusing on various units of analysis as seen above. However, in each approach some form of negotiation plays a central role in the learning process. In order to design computer support for negotiation in collaborative learning, it is necessary to specify an appropriate concept of such negotiation.
The concept of negotiation as voting seems inadequate
for CSCL. In particular, the negotiation of what is to count as new shared
knowledge for a group engaged in collaborative knowledge building has different
characteristics from other forms of group decision making. Such negotiation
might be called “knowledge negotiation” because it is not just a matter of
selecting among alternative existing states (propositions, proposals,
activation functions), but of constructing new knowledge through collaborative
interaction and discourse. The new knowledge is typically represented by or
embodied in a shared “knowledge artifact,” such as a concept, theory, text or
folder of structured information.
There is an important theoretical difference concerning the unit of analysis. We conceptualize knowledge negotiation as a group knowledge building process, rather than as a process involving individuals and their personal opinions. In CSCW negotiation, such as Herrmann’s model, commenting on one’s voting serves the purpose of expressing one’s supposedly pre-existing opinion. In BSCL, engaging in negotiational knowledge building is participating in a group reflection on shared knowledge. This difference can be seen in the thinking types of the notes contributed. In CSCW systems like BSCW, the note format stresses who the author is and may characterize the note as a “pro” or “con” opinion; in BSCL the note must first of all be determined to be a particular aspect of the group’s knowledge building process, such as a problem statement, a working theory or a summary statement before a student can begin to construct a note. Knowledge negotiation is thereby explicitly structured as part of a collaborative group effort.
Knowledge negotiation is at heart quite different from
voting. It is, in its paradigmatic forms, a nuanced give-and-take, whose aim is
to reach a solution that did not already exist in any participant’s opinion,
but that is ultimately made acceptable to all. It often involves compromises,
whereby one participant gives way in part to another’s wish in order to get the
other to give in partially to one’s own position. Negotiation is a way people
respond to non-routinized, “wicked” or ill-defined problems – where reaching
agreement often involves re-framing the issues. (Rittel & Webber,
1984)
The negotiation process as bargaining is not well modeled as a series of pre-existing positions, among which the group must vote. Nor is it well modeled as a series of positions and counter-positions among which the group must choose. In a negotiation process, typically multiple starting positions interact and evolve through a series of changing alternatives until a single consensus position is reached through discourse. The discussion is a subtle political interaction that brings many aspects of power, motivation and persuasion into play; it is a sophisticated linguistic process that cannot be algorithmically interpreted. In the end, when a consensus is reached (or not), there is often little need for a vote because agreement (or agreement to disagree) has already been established. The purpose of a vote would be to signal within a support system that everyone agreed that a consensus had been reached.
Negotiation may be conceptualized as a much broader
phenomenon than the process of making a joint decision about pre-specified
actions (or explicit access permissions). Collaborative knowledge building,
itself, can be viewed as fundamentally a knowledge negotiation process.
Proposed statements of knowledge by individuals are subjected to collaborative
interactions, whereby meanings of terms are clarified, alternative related statements
are compared, linguistic expressions are refined, warrants are scrutinized,
etc.. (Stahl, 2000)
Through these activities, the original suggestion is
transformed; through broadening consensus, the resultant expression
increasingly takes on the status of socially established knowledge. (Stahl, 2002a) Simultaneously, this process
establishes a “common ground” of understanding concerning the meaning of the
accepted expression and its constituent terms. (Baker et al., 1999) This does not necessarily
mean that every individual involved fully understands and accepts this common
ground in his or her own mind, but rather that a group understanding has been
established in the discourse of the community in which this knowledge is
thereby accepted. The co-constructed knowledge is often embodied in some form
of cultural artifact, such as a text or slogan; the common ground provides a
basis for the meaning that the artifact encapsulates to be understood in a
shared way by the collaborative community.
The shift to understanding group interactions in more dialogical terms as co-construction within a discourse community has implications for the design of groupware: away from automated selection among alternatives, toward greater emphasis on supporting communication among system users. Accordingly, it is necessary to design an appropriate mechanism for the support of knowledge negotiation in situations of collaborative learning along these lines.
BSCL is an adaptation and extension of the BSCW system
for collaborative learning applications in schools. It assigns roles of teacher
and student, that define the available functionality and access rights of the
users. Courses are usually split into smaller workgroups (typically comprising
about 3 to 7 students) that pursue specific learning goals and produce group
products or portfolios.
Each student, workgroup and course has an associated “virtual learning place,” i.e., a folder in which information and ideas are collected, typically in the form of documents, notes, links to Web pages and discussion threads. Learning places may be hierarchically structured in sub-folders. The default structure of learning places supports the concept of perspectives: There are personal, workgroup and course perspectives for students collaborating in workgroups within larger academic courses. Teachers and students can use BSCW operations to create other kinds of folder structures, but the structure to support typical workgroup collaborative activities is generated automatically by BSCL as the default.
For the knowledge building process, students typically
collect information and ideas for a learning project in their personal or group
learning places. They share and discuss these in the group learning place. The
essential task of a workgroup is to produce a group report or “knowledge
building portfolio” from collected materials and the associated discussions,
and place the report and related materials in the course learning place for
students from other groups to view and discuss.
Within an academic setting, such a contribution to the
course learning place may count as the group’s final product or work portfolio,
displayed as the group’s knowledge, shared with the other course members so
they can learn from it and comment on it. It may also be evaluated by the
teacher or others once it has reached this stage.
In BSCW, any user would be able to copy objects from a group
to a course learning place. Because of the requirements of the school setting,
it is important that a workgroup has reached a consensus on what may count as
(and be evaluated as) their group product. This requires a negotiation
function.
In a CSCW system, access rights and access functions may be
specified to an arbitrary degree of precision. This determines whether a given
user can execute a given operation under various conditions – or in BSCW it
determines whether the operation appears on that user’s menus. The rules
governing access may even be adaptable so that a group or manager can adjust
these rules. However, once set, the rules arbitrate group conflicts silently
and invisibly. For instance, if one member of a group workspace wants to delete
or edit a document and another member does not want this to happen, then the
rules determine whether it can be done or not – but the conflict between the
members who do and do not want the operation to be executed is never made
apparent. In a given case, no one knows who favors what or if and when there is
a conflict of desires, let alone people’s reasons. The systems of Herrmann and
Wulf have the advantage of making such conflicts visible and providing means
for resolving them interactively.
We were primarily concerned with transitions of knowledge from the group perspective to the course perspective. Here we wanted to bring to light any conflict within the group about promoting a knowledge artifact to the class perspective as a product of the group.
The discussion process within a workgroup may already
be considered as an implicit knowledge negotiation process. However, in the
BSCL system we make this process fully explicit to the users by commencing a
formal negotiation when a member of a workgroup proposes to promote a group
knowledge artifact to the corresponding course perspective.
Operationally, the difference between the CSCL knowledge
negotiation that is proposed here and a voting approach is that the real negotiation action is in the
evolution of the knowledge artifact proposed for agreement, and not in the
voting process itself. What is needed is to allow a proposed knowledge
artifact to be successively changed by the negotiating parties until all (or a
substantial majority) of them agree that the object is now an acceptable
representation of the group knowledge. This knowledge negotiation process may
proceed as follows:
A member of the group proposes that a specific knowledge
artifact (a set of folders, documents, ideas, or threaded discussion) be promoted
to the course perspective. Criteria for the acceptance of the proposal (e.g.,
agreement by 74% of the group within two weeks) has already been set by the
teacher for the whole class.
The knowledge artifact is made available for all group
members to modify – i.e., the object proposed for negotiation has group access
rights – within a negotiation interface at the group perspective level.
A threaded discussion area is made available for the group
members to negotiate changes to the artifact, including the statement of
reasons and suggestions for acceptable modifications.
At any point, a member can vote to accept or reject the
artifact in its current state. These votes can be withdrawn at any time, e.g.,
when a group member has made a counter-proposal which is considered more
appropriate or as the knowledge artifact is modified.
When the preset criteria for acceptance are met, the
artifact is automatically published in the class learning place. There is a
time limit for group approval; however, this is often moot since the group is
usually strongly motivated to agree on final knowledge products in order to
produce their portfolio and complete their work assignment.
In this approach, the voting interface can be extremely simple – for instance a button for the current user to signify agreement with the current version of the proposed artifact. The important point for the knowledge negotiation process is the possibility for a participant to state his or her reasons for withholding agreement in terms of dissatisfaction with the current state of the knowledge artifact. Thus, an adequate interface for the negotiation dialog is needed, in which students can formulate, exchange and react to disagreements so that the knowledge artifact can be modified in a direction that is likely to promote consensus. The knowledge negotiation interface therefore includes its own threaded discussion. When students vote, they must provide statements explaining their vote; these statements are automatically incorporated in the discussion forum where they can be further discussed. At the conclusion of negotiation, this threaded discussion represents the history of negotiation and implicitly reflects changes that have been made to the knowledge artifact as part of the knowledge negotiation, including the rationale. It can also include summary statements or a minority opinion, for instance.
The implementation of negotiation in BSCL is intended
to allow teachers to define course learning spaces that contain only shared
knowledge. The knowledge in this area is contributed by groups as such, not by
individuals. It comes from group learning spaces and represents a consensus of
the thinking of the members of that group. In the course area there is a threaded
discussion area where all course members can reflect upon the group portfolios
and build further shared knowledge on that basis at the level of the course as
a whole.
In many cases, such a strict regulation of contributions
will prove impractical and cumbersome. Therefore, we have given teachers
certain powers to short-cut or over-ride the negotiation procedures. Most
importantly, a teacher can cause a proposal folder to be published to the
course without waiting for the voting threshold to be reached. For instance,
the teacher might conduct a face-to-face negotiation with the group and then
publish the folder on that basis. Going in the other direction, a teacher might
feel that an already published portfolio has not been carefully enough
discussed and refined, and send it back for more work and renegotiation. The
teacher can also change the voting threshold for contributions to a course.
Within BSCL, teachers generally have the right to copy items between folders, including student work from group folders into course folders. However, when they do this using the new negotiation functions, the copied items are clearly marked as having been moved by the teacher from the specific group, incorporating the CSCW principle of visibility recommended by Herrmann and Wulf.
Knowledge negotiation is a central process within
collaborative learning. However, most software systems for learning do not
support it explicitly and adequately. There is a need for empirical exploration
of negotiation support functionality. BSCL will be tested in 50 schools in
The current design of support is based on our experience with the use of BSCW under many conditions during the past seven years, (Appelt & Klöckner, 1999) with studies of prototype systems by Herrmann and his students, (Herrmann, 1995; Herrmann et al., 1996; Herrmann & Kienle, 2002; Lepperhoff, 2001) and with systems created by Stahl and his associates,