It might seem logical to presume that any improvement in safety will be sufficient to convince society to adopt autonomous technology on a mass scale, but early investigations into public perception suggest that we want safety to be enormously improved before we accept driverless cars on our roads.

A recent study, published in Risk Analysis, attempts to delve deeper into this fascinating topic.  The authors set out to examine the socially acceptable risk associated with autonomous vehicles.  The results chime with previous works and suggest that we won’t accept the new technology unless it’s four to five times as safe as human-driven cars.  What’s more, people were not willing to accept safety levels comparable with human-driven cars, even with the numerous other benefits autonomous vehicles offer.  The results suggest that we significantly increase our safety demands when control is entrusted to an external force.

Safer roads

This issue is important as achieving safer roads is one of the core goals of autonomous technology.  It’s widely believed that human error contributes to roughly 94% of traffic incidents.  Whilst autonomous vehicles offer the potential to significantly improve this figure, it’s unlikely they will ever be entirely safe.  Indeed, waiting for the technology to be 100% safe is neither economically or technologically feasible.

Participants in the study were asked to complete a road safety survey, with half doing one on autonomous vehicles, and half on regular, human-driven vehicles.  Safety risk was expressed as either one fatality per x number of kilometers traveled or one fatality per x number of people in the population.  Participants were then asked to accept or reject each risk scenario on one of four levels: never accept, hard to accept, easy to accept and fully accept.

Across the board, the results suggest that people need autonomous vehicles to be up to five times as safe as human-driven vehicles.  At current levels, the traffic fatal risk is around 350 times higher than most respondents believe is required for autonomous vehicles.

Acceptable risk

The authors believe that their findings are crucial to how the technology is developed.  They propose that autonomous vehicles have their risk safety categorized according to the measures used in industrial safety.  This utilizes three criteria of risk: unacceptable, tolerable and broadly acceptable.

When autonomous vehicles are less safe than human drivers, this would be considered unacceptable.  Tolerable risk would be when autonomous vehicles are four to five times as safe (so they would reduce traffic fatalities by up to 80%).  Broadly acceptable risk would be when risks are two orders of magnitude lower than current traffic risk levels around the world.  This would provide a hundredfold improvement in road safety, or equivalent to that found in transportation modes such as rail and commercial aviation.

“Our results and method may help government authorities to establish clear safety requirements for regulating SDVs and also help SDV manufacturers find consumers’ expectations for SDVs that must be met,” the authors conclude.