Get Started
Home
Topics
Search
Library
Research questionCan causal fairness constraints transfer across synthetic-data generators and privacy levels without sacrificing fidelity?Synthetic data releases must suppress unfair causal pathways while retaining enough statistical structure for downstream use. It is unclear whether these controls remain effective when the generator family or formal privacy guarantee changes.
AI
Diffusion Models
Evaluation & Benchmarks
Machine Learning
Statistical Machine Learning
Latest papersRecent research connected to this question, newest first.Portable Causal Fairness Across Synthetic Data Generator FamiliesThe evidence covers three causal fairness definitions implemented as graph edge cuts across nine generators from marginals-based, GAN, and diffusion families, including differentially private variants at three privacy levels. Experiments use matched runs on Adult and COMPAS; the reported results indicate that the mechanism transfers across generators, with little fidelity change, a downstream classifier AUC reduction of about $0.07$–$0.15$ on average, and no observed reduction in fairness from adding privacy guarantees.research paper · Sep 2, 2026
Related questions
How can federated LiNGAM causal discovery remain reliable under near-symmetric noise without centralizing data?Can tabular foundation models learn transferable physical laws with units and noiseless mechanisms, not just interpolate data?How can complex causal queries be identified under non-IID data, transfer, and missing observations?How can differentiable causal discovery override wrong edge priors while preserving directional identifiability?