Research map / Probabilistic and scientific modeling / Treatment effects
Discovering Heterogeneous Causal Effects in Relational Data
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
A doctoral consortium thesis summary on discovering heterogeneous causal effects in relational data under arbitrary peer-influence contexts, reviewing three completed works and planned directions.
Problem
Causal inference in relational data must handle interference (a unit's outcome depends on others' treatments or outcomes) and heterogeneous peer influence, which existing methods address only for homogeneous influence or specific contexts such as neighborhood structure, roles, or groups.
Approach
Proposes a Network Structural Causal Model (NSCM) extending relational causal models with relationship existence, latent variables, and selection bias, plus a Network Abstract Ground Graph (NAGG) for identifying heterogeneous effects under arbitrary network interventions, with GNN-based representation learning of relational variables for individual effect estimation.
Key results
- The third completed work (Adhikari and Zheleva 2023) reports that ignoring heterogeneity of peer influence introduces bias and that the proposed GNN-based framework robustly estimates individual effects in networks.
- The second work shows that an ensemble of causal structure learning algorithms with ensemble-derived confidence measures gives robustness in discovering preventive and risk factors for repeat emergency room visits.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| ignore heterogeneous peer influence | individual effect estimation in networks · networks | estimation bias | Hurt |
| control for heterogeneous peer exposure using the proposed framework | individual effect estimation in networks · networks | robustness of individual effect estimation | Helped |
Limitations
- Experimental details, metrics, and baselines for the reviewed works are not reported in this summary.
- Planned interference and total-effect estimation and subpopulation discovery are anticipated work, not completed results.
Experimental setup
Datasets: Twitter public opinion data medical insurance claims data
Builds on: relational causal model Abstract Ground Graph structural causal model
Closest papers by meaning
- Do Finetti: On Causal Effects for Exchangeable Data (NeurIPS 2024 · similarity 0.87)
- Federated Causal Discovery from Heterogeneous Data (ICLR 2024 · similarity 0.86)
- On the Recoverability of Causal Relations from Temporally Aggregated I.I.D. Data (ICML 2024 · similarity 0.85)
- Discovering Mixtures of Structural Causal Models from Time Series Data (ICML 2024 · similarity 0.85)
- Telling Peer Direct Effects from Indirect Effects in Observational Network Data (ICML 2025 · similarity 0.85)
- Learning Mixtures of Unknown Causal Interventions (NeurIPS 2024 · 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.