Researchers propose a framework for planning in environments where conditions change over time. They use Partially Observable Markov Decision Processes (POMDPs) and causal knowledge to represent these changes as interventions on the environment. This allows them to evaluate plans under different scenarios and identify which parts of the environment have been altered. The approach maintains the tractability of planning methods, even when conditions change. --- Why it matters: This work matters because it addresses a common challenge in AI: adapting to changing environments. Engineers working on autonomous systems or decision-making algorithms will be interested in this framework's ability to handle distribution shifts and maintain efficient planning methods. Source: https://arxiv.org/abs/2602.23545