Causal Disentangled Anchor Learning for Scalable Fair Multi-view Clustering

Existing fair multi-view clustering methods typically suffer from a severe trade-off between clustering utility and fairness, while incurring prohibitive quadratic complexity on large-scale datasets. To address these challenges, we propose Causal Disentangled Anchor Learning (CDAL), a novel framework that achieves scalable fairness via structural disentanglement. Guided by a structural causal model perspective, CDAL utilizes a dual-anchor mechanism to structurally separate latent representations into orthogonal semantic and sensitive subspaces. We further ensure statistical independence through a linearized Hilbert-Schmidt Independence Criterion (HSIC) constraint, which is optimized via an efficient alternating scheme. Theoretically, we prove the identifiability of the disentangled factors and guarantee the algorithm's global convergence and linear time complexity $\mathcal{O}(n)$. Extensive experiments on large-scale benchmarks demonstrate that CDAL outperforms state-of-the-art methods, achieving a superior utility-fairness trade-off. Our code is publicly available at https://github.com/Tracesource/CDAL.

Machine learning algorithms used to group data for sensitive decisions (like hiring or medical diagnosis) often unintentionally reproduce societal biases. Traditionally, making these algorithms fair either ruins their accuracy or requires too much computing power for massive, real-world datasets. Instead of trying to suppress biases after the fact, our method separates the data into two isolated channels from the very beginning: one for useful grouping information, and one for sensitive bias. We also introduce a mathematical shortcut to guarantee these channels remain completely independent without slowing down the system. Our findings show that it is possible to break the usual trade-off between fairness and accuracy. This approach allows organizations to rapidly categorize large datasets while reliably filtering out discrimination, ensuring more trustworthy AI.