From Forecasts to Interventions
A forecast describes a possible outcome. We explore what additional structure and evidence are needed to reason about how an action could change it.
Two different questions
Predicting what might happen and understanding what an action might change require different kinds of evidence. A forecast alone does not establish the effect of an intervention.
We explore how temporal representations and candidate causal structure can be used together to make that distinction explicit.
Comparing possible actions
Decision support should expose assumptions, competing explanations and uncertainty. A useful comparison makes clear which conclusions depend on observed evidence and which depend on a model of the system.
Testing the effect of an action
Where intervention data are available, we would compare expected effects with observed outcomes. Where they are not, assumptions, competing causal explanations and limits of identification must remain explicit.