Analysing a repository of design knowledge with the GO-DKL browser

Abstract

This paper presents the GO-DKL browser, a web-based interface designed to support the contextualization and analysis of items within Goal-oriented Design Knowledge Libraries. It is still at a very early stage of development, though a small sample of information systems designers evaluated a demonstration of the tool and provided generally positive feedback.

Keywords

• Design Knowledge
• User Interfaces
• Knowledge reuse
• Open OME

-

2

Primary Features

Essentially, the GO-DKL browser is a web-based interface to a MYSQL database. This interface is structured to support the analytic processes discussed above, through goal association, browsing, selection, and model exploration and reconfiguration phases.

The model exploration and reconfiguration phases are centered around a tree-list interface generated using JavaScript. The tree-list structure allows relationships to be expanded and collapsed at will so that certain components can be isolated. HTML form elements permit the reconfiguration of contribution relationships. Parent goals are automatically coloured based on their received contributions from child elements, using simplified i* evaluation concepts, based largely on the propagation rules defined in [ 2 ].

The list is populated by the designer’s selected goals and any other knowledge base items that contribute to those goals (his or her ‘project model’). The actual list presentation is generated by a script that converts tabular records from the knowledge base into hierarchical HTML elements, with high-level goals as parent elements, and design features as child elements1. Often the same design feature or relationship will appear numerous times throughout the model, in which case the interface treats all instances of a it as one data point.

Therefore, if one instance of a design feature is deselected, all other instances of that feature will be deselected throughout the tree-list. In this manner, if an analyst selects a design feature because of its contribution to a focal goal, and similarly deselects another, he or she may then examine other goals to see the effects of those choice in other parts of the model. 3

Limitations, Evaluation and Future Work

While the tree-list view can help an analyst to focus intently on the contributions to a single goal (which can be difficult to do with complex models), the ‘bigger picture’ may be harder to assess. To mitigate this limitation, each high-level goal in the tree-list features an option to export a model slice[ 3 ], which retrieves all design features contributing to that goal, in addition to all other goals affected by those design features. In this manner, an analyst may examine the larger context of design decisions within his or her project model. The model slice is dynamically exported as a Q7 file, which is interoperable with Open OME2.

Several system design practitioners were shown a demonstration of the GODKL browser. They generally deemed it to be a valuable way of retrieving and informally assessing the findings of related projects in comparison to their own endeavours, though several usability critiques emerged. Future work would address these criticisms, as well as attempt to better interoperate with other goal modeling tools. 1 A maximum of 5 levels of decomposition is currently supported. 2 Importing Q7 files is currently unsupported.

Availability

Source code3 and a high-level description4 are currently available; the community is invited to contribute and expand on this work. 3 https://github.com/andrewhilts/Go-DKL also features a detailed README 4 http://www.designknowledge.org

References

1. Esfahani , H. , Yu , E. : A Repository of Agile Method Fragments . In: Mu¨nch, J. , Yang , Y. , Scha¨fer, W. (eds.) LNCS 6195: New Modeling Concepts for Todays Software Processes . pp. 163 - 174 . Springer, Berlin / Heidelberg ( 2010 )

2. Horkoff , J. , Yu , E. : A Qualitative , Interactive Evaluation Procedure for Goaland Agent-Oriented Models . In: CAISE Forum. CEUR-WS.org , vol. 453 . pp. 19 - 24 ( 2009 )

3. Leica , M.F. : Scalability concepts for i* modelling and analysis . Master's thesis , University of Toronto ( 2005 )

4. Pastor , O. , Estrada , H. , Mart´ınez, A.: Strengths and Weaknesses of the i* framework . In: Yu, E. , Giorgini , P. , Maiden , N. , Mylopoulos , J . (eds.) Social Modeling for Requirements Engineering . pp. 607 - 643 . MIT Press, Cambridge, MA ( 2011 )