A Dynamical View of the Question of Why
This addresses the problem of causal inference in dynamic systems for researchers and practitioners in fields like AI and statistics, though it appears incremental by extending existing methods to time series.
The paper tackles causal reasoning in multivariate time series data from stochastic processes by proposing a learning paradigm that directly establishes causation between events over time, using reinforcement learning to compute causal contributions. In experiments, the framework successfully reveals and quantifies causal links in intricate settings.
We address causal reasoning in multivariate time series data generated by stochastic processes. Existing approaches are largely restricted to static settings, ignoring the continuity and emission of variations across time. In contrast, we propose a learning paradigm that directly establishes causation between events in the course of time. We present two key lemmas to compute causal contributions and frame them as reinforcement learning problems. Our approach offers formal and computational tools for uncovering and quantifying causal relationships in diffusion processes, subsuming various important settings such as discrete-time Markov decision processes. Finally, in fairly intricate experiments and through sheer learning, our framework reveals and quantifies causal links, which otherwise seem inexplicable.