One way would be to use "known suspicious persons" as training material. Build the system used by the agent so that it does not immediately notify when the pass

One way would be to use "known suspicious persons" as training material. Build the system used by the agent so that it does not immediately notify when the passport for such person is scanned, but instead let the border guard do the chatting and only indicate a bit later in the process this person should not be let through. A bit similar method is (or was) used to keep the X-ray machine operators at airport awake. Software superimposes images of prohibited items on the X-ray image [1] and operator is supposed to notice them. The other case (false positives) would be harder to tackle on the case or border agent. Maybe it would help if the same agent was actually involved in the detailed investigation and not just responsible of flagging suspicious people. [1] http://www.nature.com/news/2004/040116/full/news040112-15.html http://www.nature.com/news/2004/040116/full/news040112-15.ht...