Discussion

---

The choice of intelligent agents is based on online students’ need for good support in a PBL context. The role these agents have to play is new, and they have to deal with three simultaneous dimensions:

1. The technical (how to work with portfolios and use the environment’s available tools)

2. The pedagogical (how to construct their own representation about a given domain and use a graphic environment to represent it)

3. The strategic (how to use and develop their own social competences to achieve their goals in the collaborative PBL context)

Educators responsible for following up details from all three dimensions would not be able to pay sufficient attention to the aim of the learning process. Intelligent agents can monitor students’ steps and, according to the knowledge models they have, inform students about procedures the students are not yet used to. The following examples illustrate how the agents can be intelligent:

1. In order to have an idea in the group portfolio, a student has to submit it to the other group members. The criteria used by them can be the number of arguments the idea has and the number of stars each argument has (see Figure 1). The portfolio agent can identify the patterns of “accepted” and “refused” ideas, analyze ideas in individual portfolios, and suggest that their owners better prepare these before submitting them to the other group members.

2. Idea negotiation is done by customized forms sent by e-mail. The negotiation agent can observe that a student had good ideas but used sharp words to defend them, so they were not accepted into the group portfolio. The agent informs the educator that he or she should interact with the student and prepare him or her for the next negotiation phase.

Both cases above illustrate the kind of proactiveness that is coded in models that, when fulfilled, trigger actions. Another and more powerful example of proactiveness occurs when an agent monitors the message flow and catches messages that are not addressed to it but rather to other agents related to the project it is working on. By doing so, the agent collects information that helps it to better represent the context and act in accordance with its goals than would be possible otherwise. For instance, an educator asks the assessment agent about a student’s performance in the group portfolio. The answer is that the student “seems lazy”: he or she has no ideas inside the group portfolio. The student’s interface agent, aware of that answer, sends a message to the educator, informing him or her that the student in question has evidence of good performance in his or her individual portfolio.

Our use of the SAAS method enables us to determine a primary set of services that helped the group design to a preliminary model of what a collaborative online learning environment should be like. As SAAS is a projective design method, it helped our team in its incremental brainstorming approach, which improved our confidence in the results we could achieve.

Currently, the Individual Portfolio, the Group Portfolio, the Librarian, and the Dictionary agents are being implemented in our COLE. These portfolios and agents will be operational by November 2003. Afterwards, we will be able to test the pedagogical and sociological concepts that have driven our research. Our COLE will be tested by using students in Master of Science courses and in a 40-hour lifelong learning course.

Even if the Activity Theory and the Communities of Practice model seem to be good theoretical frameworks on which to base the development of professional and social competences, we cannot comment on them until they have been tested experimentally.

We believe that our decision to use a multiagent approach will be justified as time passes, because this approach enables a system to evolve as new technologies are developed. Thus, new services with new learning capabilities and new forms of proactiveness will facilitate the activity of learning based on projects.

---