US finds border AI kiosk system too slow for operational use
In 2011, the US Department of Homeland Security funded research into a virtual border agent kiosk called AVATAR, for Automated Virtual Agent for Truth Assessments in Real-Time, and tested it at the US-Mexico border on low-risk travellers who volunteered to participate. In the following years, the system was also tested by Canada's Border Services Agency in 2016 and the EU border agency Frontex in 2014. The research team behind the system, which included the University of Arizona, claimed the system, which used sensors, biometrics, and AI to analyse movements and changes in voice, posture, and facial gestures to flag individuals who might be lying, had a success rate of 60 to 75% and was consistently more accurate than humans. DHS found, however, that the system was too slow for operational use.
Writer: Jeff Daniels