Research map / Learning theory and algorithms / Fairness

Strategic Causal Policy Learning: Welfare, Safety, and Fairness Thresholds

NeurIPS 2026 · Learning theory and algorithms › Fairness · approach: Causal and policy fairness in decisions

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Shows that the standard treat-if-nonnegative-CATE rule can fail when agents strategically manipulate covariates that determine treatment eligibility, creating a mismatch between assigned treatment and baseline welfare.

Problem

Causal policy learning typically assumes covariates used for treatment assignment are unmanipulated; when agents can move or misreport covariates after a policy is announced, treatment is assigned on manipulated covariates while welfare depends on baseline CATEs, so the standard rule can misallocate

Approach

Formalizes strategic manipulation of covariates in causal policy learning and derives policy rules (with welfare, safety, and fairness thresholds) that account for the mismatch between manipulated eligibility and baseline treatment effects, rather than thresholding CATE at zero.

Key results

• The abstract argues that the standard nonnegative-CATE rule can fail under strategically manipulable treatment eligibility; no numeric results are stated in the available abstract.

Surprising findings

• The standard nonnegative-CATE rule can fail when eligibility is strategically manipulable.

Limitations

• Full text unavailable; analysis limited to the abstract.

Experimental setup

Metrics: welfare

Baselines: treat-if-nonnegative-CATE rule

Closest papers by meaning

• Efficient and Sharp Off-Policy Learning under Unobserved Confounding (ICLR 2026 · similarity 0.87)
• Learning in reverse causal strategic environments with ramifications on two sided markets (ICLR 2024 · similarity 0.86)
• Causal Strategic Learning with Competitive Selection (AAAI 2024 · similarity 0.86)
• Towards Safe Policy Learning under Partial Identifiability: A Causal Approach (AAAI 2024 · similarity 0.85)
• A Causal Lens for Learning Long-term Fair Policies (ICLR 2025 · similarity 0.85)
• Policy Learning for Balancing Short-Term and Long-Term Rewards (ICML 2024 · similarity 0.84)