Research map / Probabilistic and scientific modeling / Treatment effects
PWSHAP: A Path-Wise Explanation Model for Targeted Variables
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use PWSHAP path-wise effects (conditional reference) instead of off-manifold Shapley for adversarial robustness | Synthetic fairness study with attacker model · Synthetic fairness study | attribution to sensitive attribute T under attacker model | Helped |
| Conditioning-based (on-manifold) reference distribution versus marginal (off-manifold) reference (ablation) | Shapley values on attacker model · Synthetic fairness study | robustness to adversarial attack (attribution to T) | Helped |
| Path-wise decomposition for local confounding effect of a true confounder | PWSHAP local bias analysis · Synthetic local bias analysis | mean |Psi|/sigma_Y for confounder vs non-confounder: 0.505 vs 0.054 when only C1 is a confounder | Helped |
| Causal Shapley direct/indirect split on bias scenarios | Causal Shapley · Synthetic local bias analysis | Causal Shapley indirect effect magnitude | No clear effect |
| Path-wise mediating effect of mediator variables versus non-mediators | PWSHAP local mediation analysis · Synthetic local mediation analysis | mean |Psi|/sigma_Y through D: 0.715 when D is mediator vs 0.041 when neither is | Helped |
| Causal Shapley indirect effect used as mediator detector | Causal Shapley · Synthetic local mediation analysis | indirect effect magnitude | Mixed |
| Local moderation analysis via PWSHAP (moderator path effects) versus Causal Shapley direct effect | PWSHAP local moderation analysis · Simulated moderation example (theoretical) | moderation attribution | Helped |
Closest papers by meaning
- Measuring Cross-Modal Interactions in Multimodal Models (AAAI 2025 · similarity 0.85)
- DISCRET: Synthesizing Faithful Explanations For Treatment Effect Estimation (ICML 2024 · similarity 0.85)
- Practical do-Shapley Explanations with Estimand-Agnostic Causal Inference (NeurIPS 2025 · similarity 0.84)
- Rethinking XAI Evaluation: A Human-Centered Audit of Shapley Benchmarks in High-Stakes Settings (NeurIPS 2026 · similarity 0.84)
- Comparing Targeting Strategies for Maximizing Social Welfare with Limited Resources (ICLR 2025 · similarity 0.83)
- Towards Causal Foundation Model: on Duality between Optimal Balancing and Attention (ICML 2024 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.