Operation Canthi

At the future-ops section of the Land Information Maneuver Center (LIMC) of the Dutch MoD, an experiment had taken place to integrate qualitative analysis by humans and deep learning. In this experiment, qualitative analysis executed by humans focused more on establishing causal relationships. The deep learning focused more on establishing correlations. In order to integrate qualitative analysis and deep learning a gap had to be bridged. The Causal Loop Diagram (CLD) served as a bridge between the two. To make this work, so-called plug-ins were introduced in the CLD. These plug-ins represented either deeper layers and/or calculations. It prevented the CLD from become unreadable complex. With the plug-ins, the CLD became de facto a three dimensional diagram. Early Warning and Early Response analysis were also integrated within the CLD by a plug-in. This calibrated version of the CLD facilitated the cross-fertilization between qualitative analysis and deep learning. Now, human analysis can form the starting point for the selection of data and indicators to be used for deep learning. Thus the number of variables used in the deep learning effort could be reduced. Also, deep learning uncovered causal relationships that went unnoticed until then. It led to regrouping of drivers within the CLD. By integrating deep learning, the future-ops section had the possibility to calculate backwards. For example, it could be calculated which measures and interventions at what point in time would be most effective. By that, it was possible to reach a maximum effect in pushing back a threat while keeping the impact for society as low as possible. In short, by calibrating the CLD, LIMC was able to merge early warning, early response, system analysis, and deep learning into a single, but layered, system analysis. It was an attempt at Augmented Intelligence – in which the cognitive process is improved and enhanced by the combination of human-machine design and assessment. Its program had never been carried out fully, as its activities were halted in November 2020.