Visualizing versus verbalizing uncertainty in intelligence analysis

Abstract

We compared the probability terms used by Western intelligence organizations against two visual encoding channel-based representations of uncertainty (i.e. darkness and thickness). Analysts were more sensitive to the probability being communicated under the word than thickness condition but not the darkness condition, with no difference among the visual conditions. However, sensitivity was not perfect. There was no difference in inter-individual variability across all conditions, which was generally poor. Test-retest reliability was greater in the word compared to thickness condition, but not the darkness condition, although, it was imperfect. Finally, analysts did not fully comply with existing uncertainty communication lexicons.