ACTION : Making Remote Sensing and Volunteered Geographic Information Decision‐Ready

ABSTRACT Remote sensing (RS) and volunteered geographic information (VGI) are routinely described as complementary, yet remain largely underexplored in the meantime. Two adjacent debates have advanced quickly: how heterogeneous spatial data should be represented within a shared learning framework, and under what normative conditions planetary‐scale artificial intelligence may be deployed. Comparatively little attention has been given to what sits between them, namely whether an integrated product changes a decision. Here we argue that RS and VGI supply different classes of evidence (i.e., calibrated measurement and situated testimony). Herein, three consequences follow this practice: (1) their combination is a judgment about evidential status rather than a preprocessing step, (2) the disagreement between them is an output in its own right rather than error to be minimized, and (3) the provenance at the level of individual assertions, not model capacity, is the binding constraint on deeper integration. We set out six conditions, abbreviated ACTION, that an ideal RS–VGI workflow shall aim to meet.