Running head: Collaborative knowledge construction FOSTERING COLLABORATIVE KNOWLEDGE CONSTRUCTION WITH VISUALIZATION TOOLS

This study investigates to what extent collaborative knowledge construction can be fostered by providing students with visualization tools as structural support. Thirty-two students of Educational Psychology took part in the study. The students were subdivided into dyads and asked to solve a case problem of their learning domain under one of two conditions: 1) with content-specific visualization 2) with content-unspecific visualization. Results show that by being provided with a content-specific visualization tool, both the process and the outcome of the cooperative effort improved. More specifically, dyads under that condition referred to more adequate concepts, risked more conflicts, and were more successful in integrating prior knowledge into the collaborative solution. Moreover, those learning partners had a more similar individual learning outcome. Collaborative knowledge construction 4 In recent years, many research activities have been directed towards analyzing discourse in cooperative learning (e. g. Dillenbourg, 1999; Mason, 1998; Kumpulainen & Mutanen, 1999). Today, we know more about discourse aspects contributing to improved learning outcomes. Scientific knowledge on processes of collaborative knowledge construction helps us to support learners more effectively in situations of collaborative learning. Recent instructional approaches include socio-cognitive structuring (e. g. O' Donnell & King, 1999) as well as shared representation or visualization techniques for fostering cooperative learning. In this paper we present findings from a study on supporting the collaborative knowledge construction with two different kinds of visualization tools. Processes of collaborative knowledge construction and cooperative learning outcomes Knowledge construction as process. "Co-construction of knowledge", "collaborative knowledge-construction", and "reciprocal sense-making" are examples of terms commonly used in research to describe the cognitive processes relevant to cooperative learning (Dillenbourg, Baker, Blaye & O’ Malley, 1995; Nastasi & Clements, 1992; Roschelle & Teasley, 1995): In theoretical and empirical papers the description or analysis of collaborative knowledge construction is often approached through the aspects of content and function of discourse. Regarding content-related aspects, a central question is to what extent, how frequently, and how adequately learners talk about the specific content of the learning task? So far, most studies have focussed on the way learners cooperatively process the content – hence, the functions of utterances in discourse are taken into consideration. For example, Renkl (1997) analyzed questions and follow-up questions in discourse, whereas Nastasi and Collaborative knowledge construction 5 Clements (1992) concluded in their research that rejection of suggestions are indicators of cognitive conflicts. Until now there has been a lack of empirical approaches which give equal weight to qualitative content-related and functional aspects. On the basis of the existing literature we distinguish four processes of collaborative knowledge construction which cover a content perspective as well as a functional perspective: (1) Externalization of taskrelevant knowledge, (2) elicitation of task-relevant knowledge, (3) conflictoriented consensus building and (4) integration-oriented consensus building. These will be described below. Externalization of task-relevant knowledge. A necessary condition for the collaborative construction of knowledge in discourse is that learners bring individual prior knowledge into the situation; only then differing views and opinions can be clarified. Especially, approaches of situated learning attach relevance to externalization, because they consider the exchange of different individual concepts to be the starting point for the negotiation of common meaning in discourse (Brown, Collins & Duguid, 1989). Also, theoretical and empirical studies generally highlight the fact that externalization is an important requirement for the "diagnosis and therapy" of misconceptions (Schnotz, 1998). (2) Elicitation of task-relevant knowledge. A further important aspect of collaborative knowledge construction is causing the learning partner to express knowledge related to the task. This is sometimes referred to as 'using the learning partner as a resource' (Dillenbourg et al., 1995). It is plausible to assume that elicitations (frequently in form of "questions") lead to Collaborative knowledge construction 6 externalizations, often in the form of explanations. Therefore, elicitations could be partly responsible for successful learning (e.g. King, 1994). (3) Conflict-oriented consensus building. Cooperative learning often causes learners to come to a common solution or assessment of the given facts. This necessary consensus can be reached in different ways. In the literature on cooperative learning socio-cognitive conflict plays an important role (Doise & Mugny, 1984; Nastasi & Clements, 1992; see Dillenbourg, 1999): It is assumed that the different interpretations made by learning partners stimulate processes which can lead to a modification of knowledge structures. (4) Integration-Oriented Consensus Building. Another way to reach consensus is to integrate the varying individual perspectives into a common interpretation or solution of the given task. This form of consensus building may be important under some conditions. However, the attempt to incorporate all individual views in a common perspective may also lead to a superficial conflict-avoiding cooperation style. The phenomenon that learners, despite drastically differing views from an objective perspective, claim that they are basically in agreement, has been observed many times (Christensen & Larson, 1993; Miyake, 1986; Hatano & Inagaki, 1991). One could speak of a tendency on the part of the learners to reach an illusionary consensus. In this study we examine, to what extent these processes can be facilitated by instructional means. We approach the measurement of these processes with combined analyses of the content and the functional level of discourse. Collaborative knowledge construction 7 Cooperative learning outcome. In research on cooperative learning there are different ideas about what is to be understood as a successful learning outcome (see Salomon & Perkins, 1998). Usually, the conditions of individual achievement are given most consideration . However, other approaches see learning as a process substantially influenced by the entire context. In this view, learning should therefore only be analyzed by taking account of the whole context (e.g. Scardamalia & Bereiter, 1994). The quality and breadth of individual knowledge construction is often given little attention in comparison to the analysis of the co-construction of knowledge in a given context. In many educational settings, it makes little sense to completely neglect the individual learning outcome, especially if not only communication and cooperation competencies are being aimed at, but also individual knowledge and skills. A second question is to what extent the individual learning partners, through cooperation, acquire similar knowledge on a subject matter. Concerning this issue, it is considered important in theoretical and empirical studies that learners negotiate a common solution, manage socio-cognitive conflicts etc. So far, however, questions about the degree in which individuals benefit differently from cooperation have seldom been raised. Does everyone learn the same amount and the same content (e. g. Cohen & Lotan, 1995)? Or does everybody learn the same amount, but in different domains, as envisioned by the concept of distributed expertise (Brown, Ash, Rutherford, Nakagawa, Gordon & Campione, 1993)? Could it possibly happen, that learners benefit from the knowledge and skill of others, without however being of any profit to learning partners? Such, usually undesirable, effects of divergence between the Collaborative knowledge construction 8 learning outcomes of learning partners are interestingly not covered in many theoretical approaches to cooperative learning. On this background we include in our study collaborative outcome measures as well as the group-to-individual transfer. Moreover, we consider intra-dyadic divergence effects. Fostering Collaborative Construction of Knowledge with Visualization Techniques An array of studies to cooperative learning has shown that efficient learning is rarely achieved solely by bringing learners together. In order for the discourse to attain a certain depth, learners usually require supportive instruction. Different forms of support for the collaborative construction of knowledge have been developed and evaluated. They often include scenarios, scripts or roles. Interestingly, most approaches are content-unspecific, i. e. they include formalisms which do not take the content of the learning environment into consideration. Through the designation of typical roles, interactive processes such as explaining and questioning are encouraged which are relevant to a vast field of content. With the goal of fostering text comprehension, reciprocal teaching is an example of content unspecific support in cooperative learning (Hart & Speece, 1998; Palincsar & Brown, 1984). This method uses the roles of 'teacher' and 'student' and can be used for supporting reading comprehension in virtually any domain. On the other hand, a more content-specific structuring method supports the learning partners in the qualitative processing of the task. In that respect, the learning partners are for example provided with an abstract diagram of the task, or a visualization of central, yet abstract characteristics of the task. In our study a content-specific visualization based on mapping Collaborative knowledge construction 9 techniques was used to facilitate the collaborative construction of knowledge. The basic principle of mapping techniques (e. g. concept mapping) is to visualize concepts (on index cards, for example) and to connect these concepts with appropriate relations. Working with such a technique results in a network (or map) of interrelated concepts. As such, a mapping technique is contentunspecific as well. However, one of the main advantages of mapping techniques for the use in cooperative learning is their adaptability to specific content. With certain types (or categories) of index cards and certain types of relations, important abstract concepts are provided that can help focus the learners' discourse on relevant aspects without undue constraint. In their content-unspecific versions mapping techniques have already proven to be effective in supporting processes of individual knowledge construction (see Jonassen, Beissner & Yacci, 1993). In particular, the acquisition of conceptual knowledge in fields like science education or preservice teacher education has been shown to benefit from mapping techniques (e. g. Beyerbach & Smith, 1990; Novak, 1998; Novak & Musonda, 1991). However, self-constructed maps proved to be more efficient than premade ones (McCagg & Dansereau, 1991). Furthermore, studies have shown that mapping techniques can, under certain conditions, also support the application of knowledge in learning with cases (e. g. Fischer et al., 1996; Mandl, Gräsel & Fischer, 2000). For several years, concept mapping has been implemented to foster cooperative learning (e. g. Plötzner, Fehse, Kneser & Spada, 1999). Initial investigations on concept mapping in cooperative learning environments indicate that it can foster a more intensive discourse between learners (Roth & Roychoudhury, 1993; van Boxtel, van der Linden & Collaborative knowledge construction 10 Kanselaar, 1997). In a pilot study from Suthers (2000) the use of a graphic mapping tool proved to be more capable of supporting cooperative learning than the textual representation. Moreover, a study by Roth (1994) showed that students emphasize the usefulness of collaborative concept mapping as a learning tool. But how can a content-specific mapping tool promote collaborative knowledge construction? We suppose that task-relevant externalization and in particular the externalization of abstract concepts as well as relations between concepts can be promoted with a content-specific mapping tool. Such a tool provides both particular categories (types of index cards) and particular relations; thus, discourse can be focused on these predicates (Collins & Brown, 1988). For example, it can be expected that collaborating on a complex problem with a mapping tool that provides the categories theoretical concept and case information will help the learner to distinguish given information or observation from interpretation on the basis of theoretical knowledge. Lacking this support, learners might use everyday concepts for the solution of a problem, without differentiating between given case information and their own interpretation. The pilot study from Suthers (2000) demonstrated that learners working collaboratively on a so called science challenge with the support of a content-specific mapping tool externalized a higher number of evidence relations than learners who where only provided with a text tool. Furthermore, a pre-structured mapping tool can help to detect missing explanations in the learners' representation: It can be seen at a glance whether a concept was used in the map, which could not be related with other cards. These "loose ends" can, then, lead to an elicitation of knowledge; a learning Collaborative knowledge construction 11 partner may happen to know a possible interpretation of a so far isolated piece of information. Two problems can arise when trying to find an adequate consensus while collaboratively solving problems. The most obvious, is that learners cannot agree on a common solution. Another possible problem is that an agreement is reached which is inadequate, because the learners have reached an illusionary consensus: Positions are taken as being mostly the same even though they are not. This may be led back to two sorts of causes: cognitive and discoursive causes on the one hand as well as emotional and motivational ones on the other hand (Christensen & Larson, 1993; Fischer & Mandl, in press). Due to cognitive and discoursive causes, the differentiation of positions held in the discussion may become more difficult to detect, for example, through a too high level of ambiguity of a claim or the lack of cognitive prerequisites (e. g. prior knowledge). If, besides that, learning partners are not motivated to cooperate, then the discourse will be held with the least possible effort (e. g. Webb, 1989). Potential conflicts will therefore be avoided. With content-specific visualization on the basis of a mapping technique, the cognitive and discoursive causes of the illusionary consensus become more apparent. We suppose that by representing the concepts and relations with the mapping technique, the ambiguity of utterances can be reduced. Differing views can be detected more easily. This possibly leads to cognitive conflict and to the negotiation of meaning both of which can be assumed to improve the learning outcome (Nastasi & Clements, 1992; Roth & Roychoudhury, 1993). To sum up, we expect positive effects of content-specific visualization on the basis of mapping techniques on the collaborative construction of knowledge. Collaborative knowledge construction 12 On this background we will examine the following research questions: (1) To what extent can processes of collaborative knowledge construction be supported by content-specific visualization? (2) To what extent can the cooperative learning outcome be improved with content-specific visualization? (3) To what extent does content-specific visualization influence (a) the group-to-individual transfer and (b) the intra-dyadic divergence of learning partners in the group-toindividual transfer?