Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Applying Attribution Explanations in Truth-Discovery Quantitative Bipolar Argumentation Frameworks

Xiang Yin, Nico Potyka, Francesca Toni

TL;DR

The paper tackles the challenge of explaining trust scores produced by cyclic Truth-Discovery Quantitative Bipolar Argumentation Frameworks (TD-QBAFs) under gradual semantics. It applies both Argument Attribution Explanations (AAEs) and Relation Attribution Explanations (RAEs), including removal-based and Shapley-based variants, to TD-QBAFs derived from truth-discovery networks using the Quadratic Energy (QE) gradual semantics. The study finds that both AAEs and RAEs yield meaningful, non-trivial explanations in cyclic graphs, with RAEs providing finer-grained, edge-level insights and revealing indirect dependencies among arguments and edges. These explanations enhance interpretability of trust and claim reliability assessments in information networks and suggest avenues for integrating multiple attribution views in complex QBAFs.

Abstract

Explaining the strength of arguments under gradual semantics is receiving increasing attention. For example, various studies in the literature offer explanations by computing the attribution scores of arguments or edges in Quantitative Bipolar Argumentation Frameworks (QBAFs). These explanations, known as Argument Attribution Explanations (AAEs) and Relation Attribution Explanations (RAEs), commonly employ removal-based and Shapley-based techniques for computing the attribution scores. While AAEs and RAEs have proven useful in several applications with acyclic QBAFs, they remain largely unexplored for cyclic QBAFs. Furthermore, existing applications tend to focus solely on either AAEs or RAEs, but do not compare them directly. In this paper, we apply both AAEs and RAEs, to Truth Discovery QBAFs (TD-QBAFs), which assess the trustworthiness of sources (e.g., websites) and their claims (e.g., the severity of a virus), and feature complex cycles. We find that both AAEs and RAEs can provide interesting explanations and can give non-trivial and surprising insights.

Applying Attribution Explanations in Truth-Discovery Quantitative Bipolar Argumentation Frameworks

TL;DR

The paper tackles the challenge of explaining trust scores produced by cyclic Truth-Discovery Quantitative Bipolar Argumentation Frameworks (TD-QBAFs) under gradual semantics. It applies both Argument Attribution Explanations (AAEs) and Relation Attribution Explanations (RAEs), including removal-based and Shapley-based variants, to TD-QBAFs derived from truth-discovery networks using the Quadratic Energy (QE) gradual semantics. The study finds that both AAEs and RAEs yield meaningful, non-trivial explanations in cyclic graphs, with RAEs providing finer-grained, edge-level insights and revealing indirect dependencies among arguments and edges. These explanations enhance interpretability of trust and claim reliability assessments in information networks and suggest avenues for integrating multiple attribution views in complex QBAFs.

Abstract

Explaining the strength of arguments under gradual semantics is receiving increasing attention. For example, various studies in the literature offer explanations by computing the attribution scores of arguments or edges in Quantitative Bipolar Argumentation Frameworks (QBAFs). These explanations, known as Argument Attribution Explanations (AAEs) and Relation Attribution Explanations (RAEs), commonly employ removal-based and Shapley-based techniques for computing the attribution scores. While AAEs and RAEs have proven useful in several applications with acyclic QBAFs, they remain largely unexplored for cyclic QBAFs. Furthermore, existing applications tend to focus solely on either AAEs or RAEs, but do not compare them directly. In this paper, we apply both AAEs and RAEs, to Truth Discovery QBAFs (TD-QBAFs), which assess the trustworthiness of sources (e.g., websites) and their claims (e.g., the severity of a virus), and feature complex cycles. We find that both AAEs and RAEs can provide interesting explanations and can give non-trivial and surprising insights.
Paper Structure (11 sections, 7 equations, 4 figures, 2 tables)

This paper contains 11 sections, 7 equations, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Preliminaries
  3. QBAFs and the QE Gradual Semantics
  4. Truth Discovery QBAFs (TD-QBAFs)
  5. Argument Attribution Explanations
  6. Relation Attribution Explanations
  7. Explaining TD-QBAFs with AAEs and RAEs
  8. Settings
  9. Results and Analysis for AAEs
  10. Results and Analysis for RAEs
  11. Conclusion

Figures (4)

  • Figure 1: Example of a TD-QBAF. (Nodes are arguments, where the $s_i$ and $c_i$ are identifiers for the source and claim arguments, respectively (for ease of reference). Solid and dashed edges indicate attack and support, respectively.)
  • Figure 2: Example of a QBAF structure for computing the QE gradual semantics.
  • Figure 3: Removal and Shapley-based AAEs for the topic argument $c5$ of TD-QBAF in Figure \ref{['fig_tdqbaf']}. (Blue/red/grey nodes denote positive/negative/negligible AAEs, respectively. The darkness of nodes represents the magnitude of their AAE values.)
  • Figure 4: Removal and Shapley-based RAEs for the topic argument $c5$ of TD-QBAF in Figure \ref{['fig_tdqbaf']}. (Blue/red/grey edges denote positive/negative/negligible RAEs, respectively. The darkness of edges represents the magnitude of their RAE values.)

Theorems & Definitions (7)

  • Definition 1: QBAF
  • Example 1
  • Definition 2: TD-QBAF induced from a TDN
  • Definition 3: Removal-based AAEs
  • Definition 4: Shapley-based AAEs
  • Definition 5: Removal-based RAEs
  • Definition 6: Shapley-based RAEs