You Have 5 Seconds To Comply

Anyone out there watch I, Robot? You know the scene where Will Smith interrogates a robot?

...actually, scratch that. You know the tabletop game Paranoia, where you're prone to getting interrogated by an insane, murderous, malfunctioning robot named Friend Computer who doesn't know the Cold War ended?

...well, let's go ahead and scratch that as well, but what I'm trying to build up to here is that we now have an interrogator robot, as the result of an experiment at Mississippi State headed by Cindy Bethel. The finding of the experiment is that the robot was less likely to introduce false information in witnesses than a human interrogator. 100 subjects were shown a slideshow of a theft, and then asked questions about it, either by a robot or a human, and split into four groups. The robot and human asked identical questions, but two groups were just asked about what they saw in the slideshow, while the other two groups got questions that introduced false information about the theft. When given false information by the human, witness accuracy dropped by 40% when compared to those given false information by the robot.

According to the report from New Scientist, it wasn't just the questions. The robot and human used identical scripts. The humans were even instructed to act as robotic as possible while asking. The bias still showed up.

The Starkville PD has already taken an interest in the results and are looking to get a robot interviewer of their own. And there's a larger implication regarding the use of robots on children, who are particularly susceptible to bias.

Not sure what the bias is towards 'all-knowing robots', though.