When it comes to bias we all think we are probably not that biased and when really pushed we can admit to some subconscious bias. This experiment conducted by Stanford reveals something much deeper and concerning. The idea that we associate black people with danger and crime in this way does raise the question about how we all conduct interviews, if these reactions are so instinctive can training, however good it is overcome some of our deeply underlying bias?
The idea that AI might at some point help overcome these challenges needs to be questioned as the premise of AI is the source of the data that allows the "machine" to make decisions without bias. What source provides the data that can be relied on to be 100% unbiased?
The article covers far more ground but the findings clearly indicate that whilst the bias exists we must all work harder to manage the bias and not excuse ourselves just because it is naturally there.
In a laboratory experiment at Stanford University, California, two groups of undergraduates were asked to look at a fuzzy computer screen. Frame by frame, out of the fuzz, an image slowly appeared: a handgun. As soon as the students could determine that it was a gun, they had to press a button. The results showed that one set of students spotted the gun significantly quicker than the other: the ones who, beforehand, had been shown pictures of black faces. The slower students had been shown white faces. When the object was not a gun or a knife but a camera or a stapler, there was no difference in reaction times. The implication was clear: the students, however subliminally, associated black people with crime and danger.
