MIT Scientists Help Self-Driving Cars To Avoid Selfish Human Drivers

Humans invented self-driving cars and we haven’t stopped trying to make it more advanced to commute effectively on the roads. When it comes to sharing space with human drivers, self-driving cars find it a lot difficult and have to face off against a slew of challenges. In particular, AI works on an algorithm-based basis that allows it to predict and interact with common human actions in rational ways. However, this becomes much more complicated for it to solve in case the self-driving car is stuck in busy traffic with human drivers.

A team of scientists from MIT has conducted a research to examine the way a self-driving car reacts to human drivers’ behaviors in the real world, which requires the prediction of high-degree social awareness. This is not easy for machines, so they took advantage of social psychology tools to make them more potent in terms of classifying behaviors of either selfish drivers or selfless ones.

In general, an observation of human behaviors when they drive will help the system to better predict sudden movements like coming to merging lanes or turning left unsafely. The new method gives it 25 percent more accurate than previously. As expected, this new insight will boost safety when there’re not only human drivers but also autonomous cars on the road.

Last year, an unfortunate accident between an Uber self-driving car and a pedestrian was due to the lack of the ability to recognize jay-walkers. The research still has room for improvement before we apply it into real life.