RALEIGH, N.C. — Researchers at N.C. State University are attempting to train self-driving cars to make the best moral decisions. 

The team has developed a virtual reality experiment to collect information from drivers about decisions they make while driving, like deciding whether or not to run a red light or go over the speed limit. 


What You Need To Know

  •  Researchers at N.C. State University are attempting to train self-driving cars to make the best moral decisions

  •  The experiment collects data via a virtual reality program

  •  Researchers hope the information will train autonomous cars to make moral decisions 

  • The team hopes to do a large-scale data collection in the future

"We have to explain not only how humans make moral judgements, but how we can train machines to make moral judgements and to see where things go wrong, even with humans,” said Dr. Veljko Dubljević, a coauthor of the experiment and professor of philosophy and science technology and Society at N.C. State.

Researchers say the data will hopefully “train” artificial intelligence to navigate different traffic situations and make sound moral decisions. 

“If you ask us if, for instance, if lying is wrong, we'll say, yes, it is. But we get into those situations where we appreciate that lying may be morally justified, such as if a serial killer asks us where some children are hiding or something like that,” Dubljević explained. “But, we have to also explain not only how humans make moral judgments, but how we can train machines to make moral judgments.”

If fully autonomous cars ever appear in North Carolina in the future, it will likely be on major roadways first. 

“I am cautiously optimistic, but I'm not sure that we're going to get there where we can trust, you know, autonomous vehicles to drive in heavily populated areas like downtown Raleigh,” Dubljević said. “So the jury's still out.”

So far, the researchers have gathered data on a small scale. 

The team hopes to soon move to the next phase of the project, which will include gathering information from thousands of drivers.