Three-way conflict analysis with preference-based conflict situations

Abstract

Existing conflict analysis models, mostly based on Pawlak's framework, start with a situation table containing agent ratings toward issues. These ratings can take various formats with differing assumptions and are often implicitly assumed to be independent. However, in practice, an agent more often specifies the ratings through relative comparisons across issues. Furthermore, consistent interpretation of ratings is hard to achieve across different agents. A numeric rating of 0.7 might indicate very strong support when provided by a conservative agent but reflect only weak support when given by a radical agent. These challenges complicate both data collection and subsequent analysis. This paper proposes a preference-based conflict analysis model to address these limitations. The model begins with preference-based conflict situations, representing pairwise preferences over issues, and defines conflict degrees based on these preferences. It further establishes three-way agent relationships to capture conflict dynamics. The model integrates seamlessly with existing rating-based approaches, demonstrated through examples involving three-valued ratings and triangular-fuzzy-number ratings. A case study illustrates its practical applicability. By prioritizing preferences over direct ratings, the proposed approach ensures more intuitive and consistent data collection while enhancing the explainability and reliability of conflict analysis.