Are autonomous vehicles safer than human drivers?

Autonomous Vehicles are driving towards zero accidents: being involved in fewer crashes, getting less injuries, and possibly saving lives. However, most comparisons between driverless vehicles and human drivers have been uneven... and maybe unfair.

Even when statistics show that above 90% of incidents in the US involve some form of human errors or deficiencies, there is not enough data to accurately assess if automation might be better than humans at not-crashing. Reliable crash rates should be calculated by incorporating how many non-collisions happen —assessing rates at which things do not happen is exceptionally challenging. For human drivers is it 1 collision per billion chances to crash? Or 1 in a trillion? To determine whether autonomous vehicles are safer than human drivers, researchers need to define a non-collision rate for both humans and self-driving cars and trucks.

As the government is allowing self-driving vehicles to operate widely, the information on fully automated systems will include more roads and situations, covering many miles in a year and as many circumstances as human drivers currently do. The crash statistics for human-driven cars are gathered from different sorts of driving situations and on all types of roads (through pouring rain, on dirt roads, climbing steep slopes in the snow, or more), and the current data on self-driving vehicles comes from good weather through unidirectional or multi-lane highways.

Both human-drivers and autonomous vehicles need to work together as driverless cars and trucks do not get emotional, tired, angry, frustrated, or drunk, but they can not foresight and avoid potential danger. Self-driving vehicles mostly drive from moment to moment instead of thinking ahead for possible events down the road; they will not react to unknown situations with the skill or anticipation of a human driver. Deciding the action to take in an emergency case is quite hard for humans and considering that autonomous vehicles have limited understanding of the world, they will never evaluate a situation the same way a human would —since machines cannot be programmed in advance to handle every potential set of events.

The promise of reducing the number of injuries and deaths is acceptable to justify expanding the use of driverless cars. However, automotive engineers need to learn from other industries (like aviation) and consider there is usually an increase in the rate of unfavorable events when implementing new automated systems in order to set up a mitigation plan. This (temporary) potential growth in the crash rate could cause concern from politicians, lawmakers, manufacturers, and even the general public —causing discouragement from supporting autonomous technologies.

Choosing to replace human-drivers with self-driving vehicles has more effects than just a one-for-one swap. Human-controlled cars and trucks are likely to remain on the roads for many years and even decades to come, so comparisons between humans and autonomous vehicles have to be carefully done. How will people and autonomous vehicles drive together? Who will be responsible in case of accidents? The need to evaluate self-driving vehicles (on how well they fulfill their promise of improved safety) is critical to ensure the data provides an accurate comparison —a fair comparison to determine if autonomous vehicles are safer than human drivers.