Building Trust in Self-Driving Cars Through Biofeedback

BraiQ wants to use bio sensors to teach you to trust your self-driving car.
April 9, 2016, 5:00pm

It has been estimated that by 2025, there will be 20 million fully-autonomous self-driving cars on the road, but one major obstacle to reaching that number remains: nobody wants to use them.

Studies show that 75 percent of Americans are afraid to ride in a self-driving vehicle. Now the people behind BraiQ, a new project from Columbia University, want to change this by teaching cars to better read human emotions. The group is developing a system to monitor human bio signals, show the car how they are feeling, and teach it how to respond.


"We think as the AI advances, unless we can develop the rapport based on the interaction to teach them to gain mutual trust, humans and machines are not going to be able to interact well," Sameer Saproo, a scientist at Columbia University and a researcher on the project said. "The autonomous car is just starting out, and the problems we are solving are going to be very prominent in the future."

The team, which includes Sameer Saproo, Sona Roy, and Victor Shih, and advisor Paul Sajda, combined their expertise in electrical engineering, bioengineering, aerospace, and brain-computer interfaces to create the technology, which they presented at the NYC Media Lab Combine demo day on Friday. The system will track biosignals of passengers including brain activity, eye movement, facial expressions, and heart rate to build a more complete picture of their psychological state. They will teach vehicle AI to better understand these emotions and respond to them appropriately.

"The idea would be that in the future when you have a self-driving car, you can actually train the car, and over a period of time it will learn what you like and don't like," Saproo said.

This means the machine would remember if you do not like fast braking or hard accelerating, and whether changing lanes is a stressor. It would be able to figure out user preferences based on specific situations –– for example, if you are napping, the car would drive more smoothly to prevent waking you, but if you are late to a meeting, the car would know it is OK to drive faster.

The technology was developed based in part on a similar project some group members were researching that created brain-monitoring technology for fighter pilots to ensure stress did not affect their accuracy. They applied these findings to self-driving technology to help cars act more naturally. Eventually, the team wants to work directly with car companies to install systems like these into vehicles.It may be awhile before companies like Ford are installing systems like these in autonomous cars, but the team has already demonstrated adaptive AI behavior in tests.

"Self-driving cars are the future, it's just a matter of how long it will take," Saproo said.

With BraiQ working to build trust and help cars act more naturally, that future could come sooner––or at least smoother.