INTELLIGENT SYSTEM FOR SUPPORTING E-PORTFOLIO-BASED LEARNING BY USING NETWORK ANALYSIS

T. Shimazaki1, Y. Morimoto1, M. Ueno2, S. Nakamura3, Y. Miyadera1

1Tokyo Gakugei University (JAPAN)
2The University of Electro-Communications (JAPAN)
3Fukushima University (JAPAN)
Introduction:
In e-portfolio-based learning, which is learning and assessment using e-portfolios, students take their assignments autonomously and communicate with peers. Therefore, e-portfolios are used as a tool to promote learning. One of the most important activities in e-portfolio-based learning is peer assessment. Peer assessment is regarded as an important assessment activity to encourage student self-assessment. In addition, e-portfolio-based learning needs learning support to enhance student learning (we call this support “facilitation”). However, it is difficult to understand the situation of peer assessment activities, for example, which student is assessed and by whom. Thus, it is difficult to facilitate each student for peer assessment activities. The goal of this study is to develop an intelligent system to understand peer assessment activities and facilitate student peer assessment. In this study, we focused on SNA (social network analysis) to achieve this goal (2). Specifically, we verified the possibility of analyzing peer assessment activities with SNA indexes (3). Next, we proposed facilitating peer assessment by using SNA (4). Finally, we developed an intelligent system that facilitates peer assessment automatically (5).

Understanding peer assessment activities:
SNA is a method that has been developed to investigate the network processes and properties of ties, relations, roles, and network formations and can analyze a whole mutual interaction made up of interconnected or related structures. Therefore, to analyze peer assessment with SNA, we defined a network structure consisting of nodes as a student and defined edges as the direction of peer assessment between students. We call this network structure PAN (peer assessment network).

Consideration of understanding peer assessment activities:
We verified the possibility of applying indexes of SNA to ones of PAN defined in the previous section. We built up a hypothesis to understand peer assessment based on PAN using each index of SNA. We had some courses in our university and collected data of the relations of mutual assessment with peers to verify the hypothesis. As a result, we found that peer assessment can be understood using each SNA index.

The SNA indexes that could be applied to peer assessment activities are as follows.

1) Degree centrality
1-1) Indegree: Students who are not assessed by others or by only a few.
1-2) Outdegree: Students who do not assess others or only a few.
2) Betweenness centrality: Students who connect to plural learning communities.
3) PageRank: Students who are assessed by students the most.
4) Cohesiveness: Learning communities in PAN.
5) Network density: Activeness of the whole of PAN

Proposal of facilitation method of peer assessment using indexes of SNA:
We proposed facilitation methods corresponding to each index of SNA verified in the previous chapter.

Development of an Intelligent System:
We developed an intelligent system that has functions corresponding to the facilitations proposed in the previous section. The system can find students who need support to be able to evaluate the situation of peer assessment in real time and provide them with accommodative facilitations.

Conclusion:
This paper proposed the way of facilitating peer assessment activities using SNA and developed the intelligence system with functions corresponding to the proposal.