Experiences and perspectives of interaction analysis
Felisa Verdejo (Moderator), Universidad Nacional de Educacion a Distancia, Madrid
Angelique Dimitrakopoulou, University of the Aegeaen, Rhodes
Agathe Merceron, Ecole Supérieure d'Ingénieurs Léonard de Vinci, Paris
Krassen Stefanov, University of Sofia, Sofia
Andreas Harrer, University of Duisburg-Essen, Duisburg

This panel compared the approaches of three Kaleidoscope projects to analysing learners' interactions. Interaction & Collaboration Analysis supporting Teachers & Students' Self-regulation, a first-year Kaleidoscope project, was presented by Angelique Dimitrakopoulou. The aim behind it was to support indiivdual or groups of students, or teachers, to regulate their activity, and also to support participants on a metacognitive level, because working in a computer-based environment, Dimitrakopoulou said, is harder than working with paper and pencil. The main goal is to develop users own regulatory capacities.

More advanced systems can evaluate collaborative modes, or perform a kind of social network analysis (similar to what David Benyon talked about at the Symposium). Indicators developed in the project describe some element related to the process of th activity, the cognitive system of an individual or community, the features or quality of a product of interaction, or the mode or quality of collaboration. As an example, Dimitrakopoulou said, in a forum discussion, a system could visualize awareness about argumentation, using correspondence analysis measured against keywords gained from data mining.

"In the ideal case," she said, "for each learing activity, we would like to have a number of indicators that when we combine them we could form a descriptive model of the interaction, appropriate for the user, in order to understand what happened." Currently there are a lot of low-level indicators, but we are lacking ones that measure cognitive aspects, and aspects of expression - for example dialogue and interaction at the same time.

This work has been extended in the current project Interaction Analysis Supporting participants in technology-based learning activities, and this was presented by Andreas Harrer. The impact of single interaction analysis tools are limited, he said, and there is a need for reusability and interoperability. Just in the few days preceding the Symposium, the project members had agreed on a common format for data. These indicators include the actions conducted by set of users, the sending and receiving ends, objects which are modified, and attributes that change; there is space for additional information like classification schema or typologies.

Regarding interoperability, he said, "It's not just only about connecting learning environments and tools, but also to combine the analysis results of tools, by piping that through.... "Right now from the talks, I know that there are a lot of connections we could do, for example with Agathe's group, with pattern mining techniques, which could be introduced in this tools suite very well." He suggested that the Kaleidoscope Shared Virtual Laboratory would be a good place to do this.

Agathe Merceron described the current project Design patterns for recording and analysing usage of learning systems, which takes a different approach to the same problem. Project members have so far compiled their own experiences in interaction analysis, and defined the state of art; now they are creating a know-how list, and beginning to structure problems for a set of design patterns. The partners' experiences were classified into the categories of course usage, collaboration, asseessment, pedagogical and administrative areas. For example, Merceron's own work attempts to get an overview of students' whole activity after they have done a set of exercises, in order to monitor and improve courses in computer programming.

Krassen Stefanov described the project's first structure for a design pattern, with five sections, each with fields or indicators. "The final product from this project," he said, "should be the design pattern browser, which actually can be used to bring the main theoretical research from this project to the public, because it can be used to represent, to share, to use, to reuse, etcetera - all experiences represented by this set of tracking design patterns."

In the discussion, Dimitrakopoulou said, "My vision of the two projects is that they are complimentary in the sense that, if you for instance are a teacher and you want to decide what kind of analysis you can use for what kind of prob you want to address, then the patterns will give you a clue about what kind of tools are there, and then you can go to the repository of tools and use one of these tools in the respository and select from there. And the interesting thing is that you use these tools that have a common format then you can also use different tools to have a different perspective on what is going on."

Harrer added that the tools could, in turn, be a source for new patterns.