1 Decision Tables for Knowledge Acquisition during Goal Interpretation P. Ardimento, M.T. Baldassarre, D. Caivano, G. Visaggio Dept of Informatics - University of Bari – Via Orabona 4, 70126 Bari - Italy; RCOST – Bari {baldassarre, caivano, visaggio}@di.uniba.it Abstract Goal oriented quality models are used in the software engineering community as a means for assessing and improving software quality. They can be seen from both a top-down and bottom-up perspective. The first for goal definition, the second for interpretation of measurement values. This paper focuses on the bottom- up perspective i.e. goal interpretation. In spite of the importance attributed to the interpretation process, literature provides little support on this aspect. The “missing link” is the lack of an “operative support” to the process itself. Interpretation process is not systematic and most likely based on personal experience and tacit knowledge of stakeholders involved. To this intent, the authors introduce Decision Tables as support to decision making during goal interpretation. Keywords: software quality, measurement, decision support, knowledge acquisition 1. Introduction Goal oriented quality models are commonly used in the software engineering community as a means for assessing and improving software process and product quality because they take into account business and project needs. Among goal oriented models we refer to the Goal Question Metrics (GQM) approach [1] which can be seen from a top-down and bottom-up perspective. Top-down for what concerns goal definition and their refinement into questions and metrics to be collected; and bottom-up for the interpretation of measurement values related to the metrics in order to answer questions and verify goal assessment. In spite of the numerous successful applications of this approach in industrial contexts [2, 3, 4, 5, 6] literature reports lacks of the approach [7, 8, 9, 10, 11] from both top-down and bottom-up perspectives. Industrial quality models also tend to be very large and therefore require much effort for definition and management (top-down). On the other hand, the fact that they include numerous goals and metrics to be measured and interpreted, inevitably increases complexity of interpretations (bottom-up). To overcome the previously listed limits related to the top-down perspective, the authors have proposed a GQM-based approach, Multiview Framework (MF) [12 13], that guides quality managers, through a set of well structured steps, in defining and managing a large goal oriented quality model. In this paper our focus is on the bottom-up perspective: goal interpretation. Although goal interpretation is considered a crucial phase of the measurement process, literature provides little support on the issue. In this sense, authors illustrate and discuss how decision tables are used as support for interpreting measurement values of goals in order to decide if and which improvement actions should be applied. Our research goals are twofold: RG1: Analyze decision tables, For the purpose of assessing their effectiveness, With respect to definition of goal interpretation, From the view point of the stakeholders, In the context of goal oriented measurement RG2: Analyze decision tables, For the purpose of assessing their effectiveness, With respect to reuse of acquired knowledge, From the view point of the stakeholders, In the context of goal oriented measurement The first research goal aims at assessing decision tables as an instrument for goal interpretation because they allow to formalize knowledge that otherwise would remain tacit and bound to the stakeholder. We hypothesize that a decision table explicates tacit knowledge making it transferable and applicable by other stakeholders. The second goal assesses that decision tables represent an important instrument for acquiring knowledge during goal interpretation. Moreover, as measurement plans are executed and interpretations are carried out, improvements are made and learning occurs with consequent modifications to decision tables. So, we hypothesize that decision tables keep track of the evolution of interpretations and of what knowledge has been collected in time. The rest of the paper is organized as follows: section 2 discusses the related literature concerning goal interpretation; section 3 introduces the reader to the general concept of decision tables; section 4 provides details on how decision tables are structured, their relation with a GQM quality goal, and how they are implied in goal interpretation; conclusions are drawn. 2. Related Literature Software engineers agree that software measurement should be goal oriented because it adapts to business and project needs. One well known approach to goal oriented measurement plan definition is the Goal Question Metrics (GQM) [1]. The main idea behind this approach is that measurement should be goal oriented and based on context characterization. It uses a top-down approach to define metrics and a bottom-up approach for analysis and interpretation of measurement data. Quality goals reflect the business strategy and goals are identified and refined based on the characteristics of software processes, products and quality perspectives of interest. Furthermore, it provides a general paradigm for defining a measurement plan. A careful analysis of literature has pointed out that much attention has been directed on aspects concerning definition of measurement goals, i.e. the top-down phase of the approach. In this sense, literature provides many examples of improvements, extensions and integrations made to the original definition of the