Research map / Learning theory and algorithms / Prediction and decision losses
Probabilistically Robust Recourse: Navigating the Trade-offs between Costs and Robustness in Algorithmic Recourse
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 |
|---|---|---|---|
| Set a user-chosen target invalidation rate and optimize for it alongside recourse cost and a positive prediction | PROBE · Adult, COMPAS, and GMC with LR and NN | average recourse invalidation rate: Adult LR: 0.35 \pm 0.11 -> 0.34 \pm 0.02; COMPAS LR: 0.47 \pm 0.02 -> 0.28 \pm 0.02; GMC LR: 0.45 \pm 0.03 -> 0.24 \pm 0 | Helped |
| Use PROBE instead of ROAR | ROAR · Adult, COMPAS, and GMC with LR | average cost: Adult: 3.56 \pm 0.8 -> 1.56 \pm 0.92; COMPAS: 2.99\pm 0.31 -> 0.63\pm 0.39; GMC: 1.74 \pm 0.45 -> 0.60 \pm 0.56 | Helped |
| Use PROBE instead of ROAR | ROAR · Adult, COMPAS, and GMC with NN | recourse accuracy: Adult: 0.94 -> 0.99; COMPAS: 0.97 -> 1.0; GMC: 0.06 -> 1.0 | Helped |
| Use PROBE instead of ARAR | ARAR · Adult, COMPAS, and GMC with NN | recourse accuracy: Adult: 0.03 -> 0.99; COMPAS: 0.02 -> 1.0; GMC: 0.06 -> 1.0 | Helped |
| Use PROBE instead of ROAR | ROAR · Adult, COMPAS, and GMC with LR | average recourse invalidation rate: Adult: 0.0 \pm 0.0 -> 0.34 \pm 0.02; COMPAS: 0.0\pm 0.0 -> 0.28\pm 0.02; GMC: 0.0\pm 0.0 -> 0.24\pm 0.01 | Hurt |
| Increase the target robustness (lower the target invalidation rate) | PROBE · Recourse cost-robustness trade-off | average cost | Hurt |
| Use a target score of 0 in Wachter recourse | Wachter · Logistic regression under Gaussian response noise | recourse invalidation rate: 0.5 | Hurt |
| Use a higher logit target score | Wachter · Logistic regression under Gaussian response noise | recourse invalidation rate | Helped |
| Use random search rather than gradient information to generate recourses | GS · GMC theoretical-bound validation | tightness of invalidation-rate upper bound | Hurt |
Closest papers by meaning
- SafeAR: Safe Algorithmic Recourse by Risk-Aware Policies (AAAI 2024 · similarity 0.90)
- Providing Fair Recourse over Plausible Groups (AAAI 2024 · similarity 0.88)
- From Search to Sampling: Generative Models for Robust Algorithmic Recourse (ICLR 2025 · similarity 0.88)
- ElliCE: Efficient and Provably Robust Algorithmic Recourse via the Rashomon Sets (NeurIPS 2025 · similarity 0.87)
- Prediction without Preclusion: Recourse Verification with Reachable Sets (ICLR 2024 · similarity 0.85)
- Algorithmic Recourse for Long-Term Improvement (ICML 2025 · similarity 0.85)
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.