How safety and reliability might influence the autonomous car’s adoption

I think it is fair to say that “safety and reliability” (such as the reduction of traffic accidents) is one of the most recurring selling points of autonomous cars (ACs)*. Although I consider the reduction of traffic accidents a weak relative advantage of the AC (relative to the manual car), I believe that “safety and reliability” encompassing, besides traffic accidents, the prevention of crime (such as car-based terrorist attacks) and recall rates will be much discussed in the autonomous car’s diffusion and will thus — paired with other factors — influence its adoption rate. How these discussions might look like and how it might affect the AC’s adoption I have explored in this post.

Less to no traffic accidents

Believing that human errors are the cause of 90% of crashes involving “passenger vehicles, trucks, and buses”, there is no doubt that the ACs could be useful here. However, considering that

  • drivers don’t care about traffic accidents
  • society doesn’t care about traffic accidents
  • drivers’ and society’s attitude is not expected to change and
  • that autonomous cars might lead to new types of traffic accidents and adverse consequences caused by temporarily higher accident rates

this alleged potential of the autonomous cars will not drive people’s adoption and might actually backfire by slowing it down.

Drivers don’t care about traffic accidents

I question the reduction of traffic accidents as an adoption driver because I have difficulties imagining that individuals perceive them as an issue. Drivers don’t care about traffic accidents because only 0,020% of all licensed drivers in the USA were involved in fatal crashes in 2015 (based on the 43,000 licensed drivers involved in “fatal crashes” relative to 218 million licensed drivers, both times in the USA in 2015).

Additionally, I don’t see any relevance for that in the context of society either.

Society doesn’t care about traffic accidents

Traffic accidents as a topic (as opposed to individual reports of traffic accidents) don’t receive much media coverage nowadays. This is one less reason for the public (politics, NGOs, the media…) to be pro autonomous cars which is of relevance because the public could not only serve as an opinion leader but also push and enable the diffusion (e. g. NGOs pressuring politics to legalize autonomous cars or the government legalizing autonomous cars).

Even worse, neither of these attitudes (public and individual interest) is expected to change in favor of the autonomous car.

Drivers’ and society’s attitude towards traffic accidents is not expected to change

Moreover, I don’t see this becoming a pressing issue in the next years neither in

  • developed countries with a saturating car market (e. g. the USA),
  • emerging countries with increased car usage (e. g. China) nor
  • “low- and middle-income” countries

Developed countries with a saturating car market (e. g. the USA)
Although there is a study from 2004 suggesting an increase in fatalities by 2020 and although the numbers have indeed been rising in the last three years, the trend has been a falling one since 2005. I believe that this trend will continue.

Emerging countries with increased car usage (e. g. China)
In other countries (e. g. China) I can imagine a temporary spike in traffic accidents in the coming years due to increased car usages. Theoretically, that issue could be solved by autonomous cars, but also — more easily — through training and education. Most of these traffic accidents will not be caused only by “unpredictable human behavior” but also by “Unsafe road infrastructure”, “Unsafe vehicles”, “Nonuse of motorcycle helmets, seat-belts, and child restraints” and “Inadequate law enforcement of traffic laws” (some of the reasons for road traffic injuries listed by the WHO). These rather predictable reasons could be solved through training and education.

This theoretical situation is, by the way, very similar to the first establishing autonomous car. It was in the 1920s when increased traffic fatalities (as a consequence of mass motorization) started receiving increasing societal attention. Since the human driver was considered fatality cause number one, removing the driver from the equation seemed like the best solution (infrastructure and car design as causes for fatalities entered the equation only at a later stage). This was when the first autonomous car — in the form of a remotely controlled driverless automobile — emerged as a response to that social problem. As we can see today the call for autonomous cars as a solution to traffic accidents didn’t bear fruits.

“low- and middle-income” countries
Although 90% of the traffic accidents happen in “low- and middle-income countries” I don’t think that autonomous cars “less accident” selling point, will be of success. On the one side, the argument mentioned for emerging countries above (traffic accidents as a natural occurrence until learning takes place) applies here as well. On the other, such countries will most probably be the last countries to adopt autonomous cars (e. g. due to financial reasons, i. e. the cars will be too expensive).

New types of traffic accidents and adverse consequences caused by temporarily higher accident rates

When reading about people’s expectations and executives promises regarding autonomous cars one gets the picture that autonomous cars will make our roads completely accident-free. Whereas I believe this vision to make sense, one should not forget that autonomous cars will bring new types of accidents with them.

New accident types
New types of accidents will be caused by the external environment as well as the internal system. In regards to the external environment, accidents will stem from the interaction between autonomous and manual cars, autonomous cars and humans as well as by all other objects unfamiliar to the car (will the car, for example, know that it can drive over the box jokingly placed in front of it?).

Internally, new accidents will stem from system failures and the car-driver interaction, like, for example, handoffs where the car and driver switch driving control. Accidents might come from the driver not being able to take over control (e. g. due to drunkenness) or the car alerting the driver to late that it is his turn for driving now (for instance, when the car is incapable of dealing with the driving conditions). Another new internal accident type would be hacked connected / autonomous cars (see below).

If such accidents do indeed happen the question is how they will influence the car’s adoption.

Accidents caused by autonomous cars could slow down its diffusion
In the transformation from horse-drawn carriages to cars the car went from being disregarded by locals to the being the road participant with the most privileges. Before the car, the street was “human domain” they were considered social gathering places with, for example, children playing**.

With autonomous cars we have a conceptually similar situation in which drivers tolerate other driver’s accidents (due to empathy) but where it is believed that humans won’t tolerate autonomous cars with an accident rate above zero (due to missing empathy; see this video).

If then autonomous cars do indeed make accidents and this empathy-assumption holds up, each accident caused by an autonomous car will slow down overall consumer adoption and possibly bringing it to a complete stop.

As opposed to new accidents types, the autonomous car might be able of preventing accidents (or rather crimes) that not only cannot be prevented by manual cars but are actually caused by them.

Preventing crimes such as car based terrorist attacks, fleeing and robbery

As using cars or trucks for terroristic attacks has become a “recommended” tactic I could imagine that such attacks will be a topic in the autonomous car’s diffusion. This assumes, of course, that car based terrorist attacks are still “a thing” by the time that autonomous car’s diffusion is being discussed in that context. Further, the prevention of crime based terrorist attacks entails also the prevention of less violent crimes such as fleeing and robbery. These topics could be framed as catalysts as well as hindrances in the autonomous car‘s diffusion. Concretely, there are three perspectives:

  1. Autonomous cars with manual driving mode and overriding privileges (capable of overriding the driver’s action)
  2. Malicious outside interference in the context of the connected car
  3. Benevolent outside interference in the context of the connected car

Autonomous cars with manual driving mode and overriding privileges (capable of overriding the driver’s action)
In this case, autonomous cars with an optional manual driving mode will be marketed as capable of detecting “terroristic driving patterns” and thus able of overriding the driver’s action when necessary. In the scenario where autonomous cars won’t have any manual driving mode at all, “terroristic driving” will be possible only through hacking and similar (see next point).

Malicious outside interference in the context of the connected car
Assuming that the autonomous car will be a “connected car” as well, discussions will arise centering around the cars’ “hackability” leading to even greater potential for terroristic attacks. Consider, for example, the case where two hackers found an exploit inside a (to the internet) connected Jeep that enabled them, amongst other things, to remotely accelerate and turn off the car (see video below).

Video demonstrating the possibilities of a remotely hacked and controlled Jeep

Just imagine “a wirelessly controlled automotive botnet encompassing hundreds of thousands of vehicles” as quoted in the article above.

Benevolent outside interference in the context of the connected car
In analogy to the above point (malicious outside interference in the context of the connected car) the cars’ connectivity feature could be controllable from some privileged official instance like the police. An example is the car robber being remotely locked into the car he stole by the manufacturer (initiated by the police). Furthermore, in analogy to the previous point, it has to be clarified that for the benevolent outside interference to be more frequent than the malicious one, the cars of the future would have to be “unhackable”.

Recall rates

According to “Wer kriegt die Kurve?” the German automobile industry’s recall rate of new cars has been rising since 2000 and cumulated to more than 50% in 2015 (see graph below).

Percentage of new cars recalled in Germany since 2000 (Source)

This means that more than every other car had to be ordered back for repair. Assuming that autonomous cars will be electric and connected cars as well, their lower mechanical complexity and possibility for over-the-air updates could result in lower recall rates and thus serve as an argument for its adoption.


*In fact, a survey by Eva Fraedrich and Barbara Lenz and published in Autonomous Driving: Technical, Legal and Social Aspects found “safety, reliability” (refers to the vision of less to no traffic accidents made possible by autonomous cars) to be number one topic of discussion among the participants. The survey analyzed 827 comments on autonomous cars made across 16 articles published in leading German and US newspapers (Germany: Bild, Die Welt, Frankfurter Allgemeine Zeitung, Heise online, Spiegel Online, Sü and Zeit Online. USA: Los Angeles Times, NY Daily News, The New York Times, San Francisco Chronicle, The Wall Street Journal and The Washington Post).

** see Technological Transitions And System Innovations: A Co-evolutionary And Socio-technical Analysis for details on that story

Leave a Reply

Your email address will not be published. Required fields are marked *