Causal-Invariant Cross-Domain Out-of-Distribution Recommendation

TL;DR

This work introduces CICDOR, a framework for cross-domain OOD recommendation that addresses co-existing CDDS and SDDS by learning dual-level causal structures (domain-specific and domain-shared) to infer invariant user preferences. It couples this with an iterative, LLM-guided confounder discovery module that extracts observed confounders from reviews, refines variables via causal discovery, and applies backdoor adjustment to deconfound recommendations. The approach demonstrates superior performance over strong baselines on real-world datasets, with extensive ablations showing the importance of both causal levels and the LLM-guided confounder pipeline. The results imply that integrating causal invariance with explicit confounder extraction can significantly improve cross-domain robustness and generalization in recommender systems.

Abstract

Cross-Domain Recommendation (CDR) aims to leverage knowledge from a relatively data-richer source domain to address the data sparsity problem in a relatively data-sparser target domain. While CDR methods need to address the distribution shifts between different domains, i.e., cross-domain distribution shifts (CDDS), they typically assume independent and identical distribution (IID) between training and testing data within the target domain. However, this IID assumption rarely holds in real-world scenarios due to single-domain distribution shift (SDDS). The above two co-existing distribution shifts lead to out-of-distribution (OOD) environments that hinder effective knowledge transfer and generalization, ultimately degrading recommendation performance in CDR. To address these co-existing distribution shifts, we propose a novel Causal-Invariant Cross-Domain Out-of-distribution Recommendation framework, called CICDOR. In CICDOR, we first learn dual-level causal structures to infer domain-specific and domain-shared causal-invariant user preferences for tackling both CDDS and SDDS under OOD environments in CDR. Then, we propose an LLM-guided confounder discovery module that seamlessly integrates LLMs with a conventional causal discovery method to extract observed confounders for effective deconfounding, thereby enabling accurate causal-invariant preference inference. Extensive experiments on two real-world datasets demonstrate the superior recommendation accuracy of CICDOR over state-of-the-art methods across various OOD scenarios.

Figures (6)

• Figure 1: An illustrative example of co-existing distribution shifts (CDDS and SDDS) in CDR.
• Figure 2: The flowchart of the proposed framework.
• Figure 3: Illustration of the prompt template for proposing candidate causal variables.
• Figure 4: Illustration of the prompt template for extracting observed confounders using zero-shot prompting.
• Figure 5: Performance comparison (HR@10) between CICDOR and baselines under different degrees of distribution shift: (a) Douban-Movie $\rightarrow$ Douban-Book OOD #1, (b) Douban-Movie $\rightarrow$ Douban-Book OOD #2, and (c) Amazon-Elec $\rightarrow$ Amazon-Cloth OOD #1.