Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

VEXA: Evidence-Grounded and Persona-Adaptive Explanations for Scam Risk Sensemaking

Heajun An, Connor Ng, Sandesh Sharma Dulal, Junghwan Kim, Jin-Hee Cho

TL;DR

VEXA addresses the need for user-facing, faithful explanations of scam risk across email, SMS, and social media in the era of fluent AI deception. It integrates detector-derived evidence via GradientSHAP with persona-adaptive explanations, forming an end-to-end four-stage pipeline that grounds language in model decision cues while modulating presentation through vulnerability-inspired personas. The paper demonstrates that evidential grounding improves semantic correctness without increasing linguistic complexity, and that persona conditioning can adjust readability and tone without disrupting faithfulness. Overall, VEXA provides a practical design framework for trustworthy, learner-facing security explanations in non-formal contexts, highlighting a layered approach where grounding governs content and persona shapes presentation.

Abstract

Online scams across email, short message services, and social media increasingly challenge everyday risk assessment, particularly as generative AI enables more fluent and context-aware deception. Although transformer-based detectors achieve strong predictive performance, their explanations are often opaque to non-experts or misaligned with model decisions. We propose VEXA, an evidence-grounded and persona-adaptive framework for generating learner-facing scam explanations by integrating GradientSHAP-based attribution with theory-informed vulnerability personas. Evaluation across multi-channel datasets shows that grounding explanations in detector-derived evidence improves semantic reliability without increasing linguistic complexity, while persona conditioning introduces interpretable stylistic variation without disrupting evidential alignment. These results reveal a key design insight: evidential grounding governs semantic correctness, whereas persona-based adaptation operates at the level of presentation under constraints of faithfulness. Together, VEXA demonstrates the feasibility of persona-adaptive, evidence-grounded explanations and provides design guidance for trustworthy, learner-facing security explanations in non-formal contexts.

VEXA: Evidence-Grounded and Persona-Adaptive Explanations for Scam Risk Sensemaking

TL;DR

VEXA addresses the need for user-facing, faithful explanations of scam risk across email, SMS, and social media in the era of fluent AI deception. It integrates detector-derived evidence via GradientSHAP with persona-adaptive explanations, forming an end-to-end four-stage pipeline that grounds language in model decision cues while modulating presentation through vulnerability-inspired personas. The paper demonstrates that evidential grounding improves semantic correctness without increasing linguistic complexity, and that persona conditioning can adjust readability and tone without disrupting faithfulness. Overall, VEXA provides a practical design framework for trustworthy, learner-facing security explanations in non-formal contexts, highlighting a layered approach where grounding governs content and persona shapes presentation.

Abstract

Online scams across email, short message services, and social media increasingly challenge everyday risk assessment, particularly as generative AI enables more fluent and context-aware deception. Although transformer-based detectors achieve strong predictive performance, their explanations are often opaque to non-experts or misaligned with model decisions. We propose VEXA, an evidence-grounded and persona-adaptive framework for generating learner-facing scam explanations by integrating GradientSHAP-based attribution with theory-informed vulnerability personas. Evaluation across multi-channel datasets shows that grounding explanations in detector-derived evidence improves semantic reliability without increasing linguistic complexity, while persona conditioning introduces interpretable stylistic variation without disrupting evidential alignment. These results reveal a key design insight: evidential grounding governs semantic correctness, whereas persona-based adaptation operates at the level of presentation under constraints of faithfulness. Together, VEXA demonstrates the feasibility of persona-adaptive, evidence-grounded explanations and provides design guidance for trustworthy, learner-facing security explanations in non-formal contexts.
Paper Structure (21 sections, 3 equations, 1 figure, 5 tables)

This paper contains 21 sections, 3 equations, 1 figure, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Scam Detection and Explainability Gaps
  4. Faithful Natural-Language Explanations
  5. Personalized Security Interventions and Vulnerability Modeling
  6. VEXA Framework
  7. End-to-End Explanation Pipeline
  8. Core Components of the VEXA Framework
  9. Experimental Setup
  10. Data Collection
  11. Preprocessing
  12. Detection Performance
  13. Experimental Design
  14. Evaluation Metrics
  15. Human Expert-Based Qualitative Evaluation
  16. ...and 6 more sections

Figures (1)

  • Figure 1: Overview of the VEXA framework. Explanations are generated through an evidence-grounded, persona-adaptive pipeline, where detector-derived cues constrain content and vulnerability personas modulate style without modeling individual users.