Exploitation and Analysis of Models

TEAM offers several ways for users to extract information from models. One way is directly provided thanks to the graphical representation of the elements of TEAM, another one is offered by session history.

First of all, the graphical representation of models reveals several aspects of the diagram. Questions solved (i.e., presenting one and only one selected option) and questions still to be resolved are quickly identified by users. This is mainly due to the different appearance of these two types of options. Poorly argumenta questions are also highlighted by the lack of criteria associated to their options. As we split design into four mains trends (related to architectural concerns) such as the dialogue part, functional core part, presentation part, and miscellaneous part, it is easy to de-tect which parts of the design have been the focus of a significant amount of work and which ones have not. The more the model contains resolved questions the more likely the model construction is close to be finished.

A lot of information can be extracted from sessions. Sessions offer a view of the diagram evolution: appearance of new branches, and disappearance of dead branches. New branches are the new questions, and dead branches are question not discussed anymore without having reached a consensus i.e., an option selected. Exploration of all sessions provides in-formation about designers’ backtracking, i.e., an option selected in one session and not selected anymore in following sessions. A regularly expanding part of a model clearly states that the question under consideration is important, gets a growing interest and raises hard to solve new problems.

DREAM provides support for constructing consistent models. All items of a duplicated option (i.e., several items sharing the same data) should have the same state: selected or not. DREAM detects such inconsistencies and notifies users. It is also possible to retrieve information (i.e., node) by query, by navigation, and visualization; these three points were defined in section ‘Retrieval Components’ in Chap 1.

DREAM provides three visualizations and two different representations of the same model. Bifocal tree representation (frame 4) modifies model representation into tree representation. Indeed, TEAM diagrams are not tree-like due to possible duplication of nodes. Leaves are factors and/or criteria. Criteria (respectively, factors) are duplicated for each relation with options (respectively, criteria). Arguments, task models and scenarios are not represented in the tree to prevent cyclic tree. A bifocal tree representation allows users to focus on a specific option without interference of other option evaluations. The bifocal tree visualization was built to support the reading activity and analyzing the importance of nodes for the design by looking (for instance) at their number of occurrence in the model.

Both representations are synchronized, a node selected in one of the view being also selected and displayed in the other views. Scalability issues (as defined in Sect. 1.7.2.) are also addressed by the bifocal tree visualization as this visualization technique has been explicitly design to handle large trees.