Research map / Reinforcement learning / MDPs and planning
Potential Outcome Rankings for Counterfactual Decision Making
Yuta Kawakami, Jin Tian
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| increase per-action sample size N' from 30 to 3000 (ablation) | PoR estimator · SCM A, K=3 | S_PoR(1,2,3) estimate vs truth 0.666: 0.557 -> 0.635 | Improved |
| increase per-action sample size N' from 30 to 3000 (ablation) | PoB estimator · SCM A, K=3 | S_PoB(1) estimate vs truth 0.666: 0.561 -> 0.650 | Improved |
| increase number of candidate actions K from 3 to 20 (ablation) | PoR estimator · SCM C (N'=3000) | absolute error of PoR: 0.022 -> 0.256 | Hurt |
| increase number of candidate actions K from 3 to 20 (ablation) | PoB estimator · SCM C (N'=3000) | absolute error of PoB: 0.015 -> 0.023 | Hurt |
| reduce per-action sample size from 3000 to 300 at K=20 (ablation) | PoR estimator · SCM C | absolute error of PoR: 0.256 -> 0.431 | Hurt |
| drop rank invariance and use bounds instead of point estimate (ablation) | PoR bounds · SCM B, K=3 (N'=3000) | upper bound of S_PoR vs true 0.500: 0.873 upper bound, 0.000 lower bound, true value 0.500 | Mixed |
| use PoR ranking instead of RoE ranking for treatment choice | treatment decision · Coagulation Thromb.count | ranking of treatments: RoE: B>H>S (0.994, 0.917, 0.873); PoR best (B,S,H) 0.277 | Mixed |
| use PoB instead of RoE to assess the best treatment B | treatment decision · Coagulation Thromb.count | share of patients for whom B is best: B is optimal for only 49% of patients | Mixed |
Closest papers by meaning
- Learning Counterfactual Outcomes Under Rank Preservation (NeurIPS 2025 · similarity 0.88)
- Decision Theoretic Foundations for Conformal Prediction: Optimal Uncertainty Quantification for Risk-Averse Agents (ICML 2025 · similarity 0.85)
- Counterfactual Effect Decomposition in Multi-Agent Sequential Decision Making (ICML 2025 · similarity 0.84)
- Counterfactual Identification Under Monotonicity Constraints (AAAI 2025 · similarity 0.84)
- Predictive Performance Comparison of Decision Policies Under Confounding (ICML 2024 · similarity 0.84)
- Designing Decision Support Systems using Counterfactual Prediction Sets (ICML 2024 · similarity 0.84)
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.
