Generalize temporal-window parameter guidance for coordinated behavior detection

Establish principled guidance for selecting temporal window parameters in coordinated behavior detection, addressing the unresolved methodological challenge of determining appropriate window lengths and strides for detecting coordinated communities across different settings.

Background

The paper studies how temporal window length and stride influence coordination networks and the communities detected from them. Using temporal co-repost networks built from labelled information-operation campaigns, the authors find that window length substantially changes which coordination patterns are observable, whereas window stride has little effect on precision and recall in the tested configurations.

Despite these empirical findings, the paper does not establish generally applicable rules for selecting temporal parameters. The appropriate temporal scale may depend on the synchronization pattern, modality, campaign, and network-construction procedure, making parameter selection an unresolved methodological issue.