Explaining Projected Answer Sets Using Faceted Reasoning
Explaining Projected Answer Sets Using Faceted Reasoning
Masterarbeit von Paramita Choudhury
- Betreuer Sarah Gaggl
- Logische Programmierung und Argumentation
- – 27. November 2025
and reasoning, widely used for modeling combinatorial problems in artificial intelligence. Modern ASP solvers can generate large numbers of answer sets, but exhaustive enumeration is often impractical due to computational constraints and the difficulty of interpreting massive solution spaces. Projection addresses this challenge by reducing answer sets to atoms of interest, but it introduces a new problem: users lack insight into what information was eliminated and whether the projected results adequately represent the underlying diversity. This thesis presents a facet-based approach to explain projected answer sets in ASP. We develop an algorithm that computes facet counts - measures of diversity based on brave consequences - for projected-away solution spaces. By formulating integrity constraints from projected answer sets and invoking facet-counting solvers, our approach quantifies how much variation remains hidden behind each projection. We formalize this through three key definitions and implement it as EPAS (Explaining Projected Answer Sets), a tool supporting bounded-time execution and interactive navigation. Additionally, the system includes an interactive feature that enables users to explore structural relationships within projected answer sets, transforming projection from a black-box reduction into a transparent analytical process.
