Discontinuity-Free Decision Support with Quantitative Argumentation Debates

The representation of decision problems is an extremely challenging task, especially when these problems are wicked, namely they rely on incomplete or conflicting knowledge, they involve several stakeholders with conflicting views and they are closely linked with other decision problems (Churchman 1967). Instances of wicked decision problems occur naturally, for example, in engineering design (Marashi and Davis 2006) and e-democracy (Loukis, Xenakis, and Tseperli 2009; Gordon and Richter 2002). IBIS (Issue Based Information System) (Kunz and Rittel 1970) provides a powerful and widely adopted approach for knowledge representation especially suitable for wicked decision problems. While many tools for visualisation and collaborative development of IBIS graphs are available (Conklin et al. 2001; Bracewell et al. 2009; Aurisicchio and Bracewell 2013), automated decision support in this context is still underdeveloped, even though it would benefit several applications. QuAD (Quantitative Argumentation Debate) frameworks (Baroni et al. 2015) are a recently proposed IBIS-based formalism encompassing automated decision support by means of an algorithm for computing the strength of answers to decision questions, given an IBIS graph of a restricted kind (see (Baroni et al. 2015) for details). The calculation aggregates the strength of attacking and supporting arguments for the answers and for other opinions, recursively. The QuAD algorithm has been integrated within the designVUE tool1 (Baroni et al. 2015) and the Arg&Dec system2 (Aurisicchio et al. 2015) and has proven useful in several applications in engineering design. Figure 1 shows an IBIS graph/QuAD framework as visualized in Arg&Dec. The graph is an abstraction of a real decision-making problem in design engineering, where the Issue node represents the question of how to control the ventilation in a room; the Answer nodes A1 and A2 are the available options, with A1=“building management system control” and A2=“user control”; the two Pro Argument nodes P1 and P2 support A1 and P3 supports A2, with P1=“energy is saved”, P2=“elderly occupants require more simple settings” and P3=“user satisfaction is increased”; the Con Ar-