Generative multitask learning mitigates target-causing confounding. (arXiv:2202.04136v3 [cs.LG] UPDATED)

We propose generative multitask learning (GMTL), a simple and scalable
approach to causal representation learning for multitask learning. Our approach
makes a minor change to the conventional multitask inference objective, and
improves robustness to target shift. Since GMTL only modifies the inference
objective, it can be used with existing multitask learning methods without
requiring additional training. The improvement in robustness comes from
mitigating unobserved confounders that cause the targets, but not the input. We
refer to them as \emph{target-causing confounders}. …

arxiv multitask learning