Georgios Kahrimanis

Assessing Collaboration Quality in Synchronous CSCL Problem-Solving Activities: Adaptation and Empirical Evaluation of a Rating Scheme

The work described is part of an ongoing interdisciplinary collaboration between two research teams of the University of Patras, Greece and the University of Freiburg, Germany, which aims at the exchange of analysis tools and data sets in order to broaden the scope of analysis methods and tools available for Computer-Supported Collaborative Learning (CSCL) support. This article describes the adaptation, generalization and application of a rating scheme which had been developed by the Freiburg team for assessing collaboration quality on several dimensions [1]. The scheme was successfully adapted to suit data gathered by the Patras team in a different CSCL scenario. Collaboration quality is assessed by quantitative ratings of seven qualitatively defined rating dimensions. An empirical evaluation based on a dataset of 101 collaborative sessions showed high inter-rater agreement for all dimensions.

On supporting users' reflection during small groups synchronous collaboration

During computer-mediated synchronous collaboration there is need for supporting reflection of the partners involved. In this paper we study techniques for determining the state of an evolving collaborative process, while the activity is in progress, making the users aware of this state. For this reason, a State of Collaboration (SoC) indicator has been defined, which is calculated using a combination of machine-learning and statistical techniques. Subse-quently a study was performed during which SoC was presented to a number of groups of collaborating partners engaged in problem-solving activities. It was found that this group awareness mechanism influenced in a significant way the behavior of the groups in which it was used. This study has wider implications to the design of groupware and in particular towards gaining an insight into the effect of group awareness mechanisms on computer-mediated collaborative learning.

A Model for Interoperability in Computer Supported Collaborative Learning

Computer Supported Collaborative Learning activities involve combination of complex software tools that often need to interoperate in a wider context of learning. This paper proposes a data model that accommodates requirements of typical collaborative learning situations and facilitates interoperability of tools and interchange of products of collaboration and evaluation data. The model has been tested against various typical tools used for both synchronous and asynchronous collaboration of groups of students.

Towards a flexible model for computer-based analysis and visualization of collaborative learning activities

The definition of appropriate interaction analysis methods is a major research topic in Computer Supported Collaborative Learning. Analysis methods can be totally or partially supported by computer-based tools that provide for better and more efficient analysis processes. The current research in this field shows that most interaction analysis tools have been based on unstable prototypes, and are highly dependant on the learning environments and research goals for which they were defined. As a consequence, it is not possible to use them in authentic CSCL settings with real users. The goal of this European Research Team therefore is to utilize the synergies of experience in manual interaction analysis with computer-based analytical methods. In this article we present an approach that embeds standardized computer-supported techniques into a semi-formal analysis process model which can be utilized and adapted in a flexible way according to the cases and environments to be analysed.

Interaction Analysis as a Tool for Supporting Collaboration: An Overview

This chapter constitutes an overview of logfile-based interaction analysis techniques that can be used for the support of Computer Supported Collaborative Learning (CSCL) activities. Interaction analysis is central in the study of CSCL activities, since in such activities through interactions between partners the state of evolving group knowledge is communicated. This interaction is facilitated by tools that allow logging of events that take place, capturing thus information about the content and the process of collaboration. Automated analysis techniques of this information can be developed. The objective of this analysis is often to support participants, in several ways: explicitly, by providing feedback to them in order to regulate their practices, or by making adaptive changes to some aspects of the collaborative setting; or implicitly, by making available to them representations of their activities. This chapter presents the most common approaches used in interaction analysis, while it particularly emphasizes recent innovative efforts to reap the advantages of machine learning techniques in order to overcome common shortcomings of previous approaches.

Validating Empirically a Rating Approach for Quantifying the Quality of Collaboration

Interdisciplinarity in the Computer Supported Collaborative Learning (CSCL) research field involves the application of several methodological approaches towards analysis that range from deep-level qualitative analyses of small interaction-rich episodes of collaboration, to quantitative measures of suitably categorized events of interaction used as indicators of the success of collaboration in some of its facets. This article adopts an alternative approach to CSCL analysis that aims at taking advantage of some desired properties of each of these diverse methodological trends, involving the use of a rating scheme for the assessment of collaboration quality. After defining a set of dimensions that cover the most important aspects of collaboration, it employs appropriately trained human raters basing their assessments on substantial aspects of collaboration that are not easily formalisable. The activities studied here regard 228 collaborating dyads, working synchronously on a problem-solving task. Based on this large dataset, relations between dimensions of collaboration quality are unraveled on empirical grounds, by elaborating ratings statistically using a multidimensional scaling technique.

Determining Relations Between Core Dimenisons of Collaboration Quality: A Multidimensional Scaling Approach

Computer Supported Collaborative Learning (CSCL) constitutes one of the most extensively developed paradigms of research and practice in intelligent networking and collaborative systems technology. Interdisciplinarity in the research field involves the application of several methodological approaches towards analysis of CSCL that range from deep-level qualitative analyses of small interaction-rich episodes of collaboration, to quantitative measures of suitably categorized events of interaction that are used as indicators of the success of collaboration in some of its facets [1]. This article adopts an alternative approach to CSCL analysis that aims at taking advantage of some desired properties of each of these diverse methodological trends, involving the use of a rating scheme for the assessment of collaboration quality [2,3]. After defining a set of dimensions that cover the most important aspects of collaboration, it employs appropriately trained human agents to assign ratings of collaboration quality to each dimension, basing their assessments on substantial aspects of collaboration that are not easily formal sable. The activities studied here regard 228 collaborating dyads, working synchronously on a computer science problem-solving task with the use of the Synergo tool [4]. Based on this large dataset, relations between dimensions of collaboration quality are unraveled on empirical grounds, based on the ratings of collaboration quality that were elaborated statistically using a multidimensional scaling technique [5,6,7,8,9,10,11]. Results obtained are in accordance with the initial design of the rating scheme used, and further particularize the relations between the dimensions it defines.