We study inference on the long-term causal effect of a continual exposure to a novel intervention, which we term a long-term treatment, based on an experiment involving only short-term observations. Key examples include the long-term health effects of regularly-taken medicine or of environmental hazards and the long-term effects on users of changes to an online platform. This stands in contrast to short-term treatments or "shocks," whose long-term effect can reasonably be mediated by short-term observations, enabling the use of surrogate methods. Long-term treatments by definition have direct effects on long-term outcomes via continual exposure so surrogacy cannot reasonably hold. Our approach instead learns long-term temporal dynamics directly from short-term experimental data, assuming that the initial dynamics observed persist but avoiding the need for both surrogacy assumptions and auxiliary data with long-term observations. We connect the problem with offline reinforcement learning, leveraging doubly-robust estimators to estimate long-term causal effects for long-term treatments and construct confidence intervals. Finally, we demonstrate the method in simulated experiments.

翻译：我们研究基于仅涉及短期观测的实验，推断持续暴露于新型干预（称为长期处理）的长期因果效应。关键示例包括定期服用药物的长期健康影响、环境危害的长期健康影响，以及在线平台变更对用户的长期效果。这与短期处理或“冲击”形成对比——后者的长期效应可合理通过短期观测介导，从而能够使用替代方法。长期处理在定义上通过持续暴露直接作用于长期结果，因此替代性假设无法合理成立。我们的方法直接通过短期实验数据学习长期时间动态，假设观测到的初始动态持续存在，但避免了替代性假设及需使用包含长期观测的辅助数据这两项要求。我们将该问题与离线强化学习联系起来，利用双稳健估计量估计长期处理的长期因果效应并构建置信区间。最后，我们通过模拟实验验证了该方法。