Training Robots & Human Bias
Robotic development is creating the construction, operation and design of these new robotic computer systems which can substitute and replicate human actions. From science fiction, to cinema and, to working next to us in the workplace, and perhaps, soon, even driving us home from the office, robots have been steadily advancing into our everyday lives.
This development is constantly going forward but often unintended human input courses problems with the robotic outcome. Ayanna Howard is Professor and chair of the school of interactive computing at Georgia Institute of Technology and a self-proclaimed “roboticist extraordinaire.”
Howard said she feels it is important for people to manage their expectations when it comes to robots as many have the view that they should be infallible.
“Somehow we believe that robots will be smarter than they are, and not have any mistakes, but how many times have you had to reset your phone or your computer? Technology is not perfect and these robots will not be either.”
Howard made clear that robots inherit flaws from the humans who create them and the data they are programmed with. “We train these robots based on human values, human concepts, human data, human understanding but it also means that all of our historical biases are put into the machine,” she said.
As a robotic programmer she collects data and analyses it for healthcare and models clinicians and doctors.
“If the historical data is wrong and I’m doing it right then my robot is doing exactly what it is supposed to do. It has that inaccuracy but if the data is wrong because we as humans have been doing things wrong, my system doesn’t know,”
This statement sounds almost like a call to get more domain expertise into the programming stage so that biased, erroneous or problematic data can be highlighted and eradicated or counteracted before it is programmed into the robot. This could happen through the roboticists embedding with the clinicians and doctors to get a thorough understanding of what good looks like and bring that back to the lab.
Alternatively, a clinician could work with the roboticists in the lab to highlight and correct any errors or mistakes that they see.
Howard also called on the public to keep roboticists and programmers in check by pushing back on questionable or problematic developments. She gave the example of one robotics company that proclaimed to come out with a robot that would teach children manners. However, there was a strong reaction from parents who made it clear manners are imbued with values and it was not the place of a commercial company to instill values in their children.
Prof. Howard believes robotics is an exciting area to work in because there is a universal fascination with robotics and everyone from the very young to very old has an opinion about robots and most has their favourite one. That could be Wall-e, iRobot or Terminator.
You Might Also Read:
Next-Gen Robotic Process Automation Leverages AI And Machine Learning: