Text-PRAS: An Explanation-Alignment Audit Framework for Arabic Stance Detection

Anonymous Authors
Anonymous Affiliation
anonymous@example.com

Abstract

We introduce Text-PRAS, an attribution-mass audit framework for Arabic target-conditioned stance detection. Text-PRAS quantifies how post-hoc attribution mass is distributed over predefined stance-evidence and shortcut/noise token categories. On the official public Mawqif-v2 train/dev split using MARBERT, raw Text-PRAS scores are positive across Integrated Gradients (IG), GradientShap, and LIME. However, within-example shuffled attribution baselines are consistently higher, yielding negative baseline adjusted deltas. This shows that raw positive attribution-mass scores can overstate explanation alignment unless controlled against Atlas/token coverage. Text-PRAS is therefore presented as an audit signal, not as proof of true reasoning or explanation faithfulness.

1. Introduction

Arabic stance detection has important applications in social media analytics, public opinion monitoring, misinformation analysis, and evaluation of public discourse. However, high stance classification performance alone does not establish whether models rely on stance-relevant textual evidence or on shortcut cues such as hashtags, URLs, mentions, repeated slogans, target-only phrases, or boilerplate text.

We introduce TextEvidenceAtlas and Text-PRAS as an explanation-alignment audit framework for Arabic target-conditioned stance detection. The goal is not leaderboard optimization. Instead, the framework asks whether post-hoc attribution mass is distributed over predefined stance-evidence token categories or over shortcut/noise categories.

Why Arabic requires a dedicated audit

Arabic social-media text presents attribution challenges that are not adequately handled by directly translating English rationale or shortcut taxonomies. Arabic is morphologically rich: clitics attach to words, negation and prepositions may be short or attached, and subword tokenization can fragment stance-bearing words. Dialectal and orthographic variation produce many surface forms for the same stance cue. Arabic-script hashtags often conflate ordinary words, target mentions, campaign slogans, and stance markers in a single token. Sarcasm, contrast, and quotation patterns in right-to-left social-media text may invert the stance implied by local sentiment words. TextEvidenceAtlas categories therefore explicitly include Arabic stance markers, negation/modality cues, contrast cues, target anchors, Arabic sentiment cues, and Arabic Twitter artifacts such as hashtags, mentions, repeated emojis, repeated slogans, and boilerplate phrases. This makes the framework task-general within Arabic target-conditioned stance detection, but not universal across Arabic NLP tasks.

Contributions

We make three bounded contributions. First, we introduce TextEvidenceAtlas, a transparent rule-based atlas for Arabic target-conditioned stance detection that distinguishes stance evidence from social-media shortcuts. Second, we define Text-PRAS, an attribution-mass audit score that measures evidence mass minus shortcut mass and explicitly reports coverage, sensitivity diagnostics, overlap diagnostics, and a within-example random baseline. Third, we evaluate the audit on MARBERT over the Mawqif-v2 official public train/dev split using three attribution methods, showing both method agreement and method dependence while avoiding claims of reasoning, causality, or universal validity.

Scope

The unseen test is pending official release. We do not claim final unseen performance, model understanding, causality, or proof of reasoning. Text-PRAS is an explanation-alignment audit signal.

2. Related Work

Arabic stance detection and social-media modeling. Arabic stance detection has been studied in fact-checking, claim verification, and social-media settings. Mawqif-v2 focuses on target-conditioned Arabic tweets labeled as Favor, Against, or None. Our work differs from modeling-centered studies: we do not aim to improve leaderboard performance, but to audit whether attribution mass for a trained Arabic stance model aligns with predefined stance-relevant textual evidence rather than social-media shortcuts.

Interpretability and attribution in NLP. Post-hoc explanation methods such as LIME, Integrated Gradients, and SHAP-style methods provide token-level importance estimates, but their outputs are not ground truth explanations. Text-PRAS is not a claim that the model understands stance; it is an explanation-alignment audit signal measuring whether attribution mass falls on atlas-defined categories.

Shortcut learning and annotation artifacts. NLP models often achieve high task performance by exploiting artifacts or shallow heuristics. Our shortcut categories are adapted to Arabic social-media stance detection.

3. TextEvidenceAtlas

TextEvidenceAtlas is a predefined rule-based atlas. It is task-general within Arabic target-conditioned stance detection, not a universal Arabic NLP metric.

Table 1. Evidence categories in TextEvidenceAtlas v1.

Table 2. Shortcut categories in TextEvidenceAtlas v1.

Overlap caveat. Atlas rules may assign overlapping evidence and shortcut labels to some tokens, especially target-related or hashtag-like tokens. We therefore report overlap diagnostics and rely on the canonical attribution-mass accounting used by Text-PRAS.

Table 3. Evidence/shortcut overlap diagnostics.

Table 4. Strict tokenization-aware Atlas coverage.

4. Text-PRAS

Text-PRAS = EvidenceMass − ShortcutMass, where EvidenceMass and ShortcutMass are computed from absolute attribution mass over Atlas categories.

5. Experimental Setup

6. Results

Raw Text-PRAS and shuffle baseline

Raw Text-PRAS scores are positive, but the within-example shuffled baselines are consistently higher. Therefore, baseline adjusted deltas are negative for all three methods. This means raw positive Text-PRAS would have overstated explanation alignment if not controlled against Atlas/token coverage.

Attribution-mass coverage

Atlas v1 sensitivity

Atlas v1 sensitivity should be interpreted as raw-score sensitivity analysis, not random adjusted evidence alignment.

7. Discussion

Raw positivity versus baseline adjusted alignment. Although raw Text-PRAS is positive for all methods, within-example shuffled baselines are consistently higher, with negative deltas whose confidence intervals lie below zero. This indicates that attribution mass is not concentrated on evidence tokens above what Atlas/token coverage alone would predict under random attribution.

Evidence alignment is not sufficient for correctness. Text-PRAS measures attribution alignment with atlas-defined evidence categories, not correctness, compositional understanding, or causal reliance. Text-PRAS should be used alongside, not instead of, task metrics.

8. Conclusion

We introduced Text-PRAS, an attribution-mass audit signal for Arabic stance detection. The finding that observed Text-PRAS falls below within-example shuffle baselines is itself a meaningful audit result: it demonstrates that the framework can distinguish between raw positivity driven by Atlas/token coverage and genuine above-random evidence alignment.

Limitations

1. Results are bounded to Mawqif-v2, MARBERT, and the official public dev split.
2. The unseen test is pending official release; no final unseen result is claimed.
3. IG, GradShap, and LIME are post-hoc approximations; none is ground truth.
4. Atlas rules may produce overlapping evidence and shortcut labels; overlap is reported diagnostically.
5. Positive raw Text-PRAS does not imply correctness, causality, true reasoning, or explanation faithfulness.

References

References should be compiled from references_expanded.bib in the ACL LaTeX version.