Self-driving Vehicles- Gearing Up to Save Lives, Reduce Costs, Resource Consumption

Introduction

During last years a lot of worldwide leader brands are investing in research of self-driving cars. The benefits of this innovation start from the possibility to reduce the number of traffic accidents on our roads by 90% (in Europe they cause more than 25,000 fatalities each year) but they wouldn’t stop here: it could increase traffic efficiency , reduce pollution and allow customers to spend time more efficiently. Each year the average American spends about 17,600 minutes driving this time could be spent relaxing, reading or even working. However, this kind of innovation must face some important barriers that could seriously delay the diffusion. From the technical side we are making great strides but there is an underline issue that is probably harder to treat: are we ethically ready for this incredible innovation?

After a generic description of the innovation and its main characteristics we will analyze the possible limits on the diffusion. Starting from the technological challenges we will end up analyzing the sneakier but probable stronger obstacle of the ethical dilemmas.

Analysis of the innovation

In order to have a more complete picture of this technology we firstly need to frame it. We are talking about a product innovation that could be both considered as incremental or radical depending on the level of interaction between the human and the car. We can think about some already existing models in which the user is only assisted by the automobile and consider them as an improvement of an already present technology, but if we take for example the model of “robo-taxi” by Google, in which the car completely controls the driving process, we can say that it’s a radical innovation because it deeply changes the concept of daily travel. In this brief research we will focus on the “Google model” because it presents some interesting problematics on the future diffusion.

This kind of innovation will require a strong level of interaction with nearby vehicle in order to increase the efficiency of this transport system. The value of the consumption does not only depend on the characteristics of the vehicle but also on the number of others self-driving car users. The more the share of autonomous vehicles (AVs) the more the positive externalities due to network effects are stronger and could be exploited; on the other hand a higher share of human driven cars would increase the risks and decrease the efficiency of the system. Imagine a world in which 90% of cars are self-driven and 10% are still controlled by humans: all AVs would be able to communicate within each-others, knowing when another car is going to curve or brake in advance so they can correct the speed and trajectory, but what about the unpredictability of the human behavior? An improvise turn, a moment of sleep or a simple distraction could cause a disaster across the lane.

Another advantage regards the optimization of traveling duration: knowing exactly the intention of nearby vehicles you could rise the speed limits on the roads since almost everything would be monitored. Also traffic lights could be removed since each car can predict exactly the intentions of others. In other words to be really efficient this innovation needs to exploit the benefits from the network effects, both direct and indirect:

• direct: as we said, the user benefits directly from the existence of large network of other users, the more the share of self-driving cars the more would be the efficient of the network and the less are the risks connected to it;
• indirect: users would also take benefits that not derive from other users directly such as more car replacement parts presents in the market.

Challenge of diffusion

Let’s see in general terms what could stops the spread of this innovation. Probably at the beginning of their lifecycle AVs would cost a lot compared to usual ones but they could target the luxurious part of the market as long as prices would stay high. There could be also other limits to the diffusion of this technology: the efficiency related to the network, risks related to cyberattacks, the psychological block related to trust and the ethical decisions that should be taken in critical situations.

As we previously discussed the efficiency of this new generation of cars is inversely proportional to the number of human-driven cars still present on the road. To overcome this problem we could adopt some partial remedies such as creating new roads that can only be traveled by self-driving cars that can communicate within each other. It follows that the more the network of users will expand the more we will need powerful technologies able to communicate and elaborate an increasing number of data. A consequent step may be merging this road to the classic ones, creating a single one with separate lanes for different kinds of cars, but the more we put closer these heterogeneous vehicles the higher the risks. Another technological weak-spot of this innovation will be the cybersecurity and, more in general, all problems related to hacking. Let’s suppose a possible scenario on a highway of the future that will be traveled by a lot of fast automated vehicles and try to imagine the tragic consequences of a pirated car, or even worse a truck, launched at full speed in the traffic.

There are also other big roadblocks that may obstacle the diffusion of this technology but they are different from the technological ones. A survey made on American drivers showed that 78% of the attendants reported fear during a travel in an autonomous vehicle and only 19% said they actually trusted the car. This is because by using this technology we completely give up the control of the vehicle, furthermore to an artificial intelligence: mankind would need some time to get used to this.

Probably all these problems would take some time to be resolved and the integration should be gradual. In the next section we would discuss a not obvious and more sneaky problem which may delay the diffusion for a lot of years, if not even interrupt it.

The ethical problem

When the human being is facing an improvise and critical situation, which require a choice to be made, there is no time to evaluate the consequences of each possible alternative so our instinct comes into play: for self-driving cars the situation is completely different.

Even in the best possible framework we would not be able to avoid all possible crashes; sometimes AVs will have to make decisions that could involve unavoidable harm to passengers or pedestrians. This judgment will be based on algorithms programmed in the car a priori by their designers that should face some serious moral dilemmas. Imagine an extreme but still possible scenario in which the AV must decide whether to sacrifice pedestrians in order to save its passenger or vice versa. There are two main way to face this dilemma: choosing the utilitarian principle by saving the most lives or acting as self-protective prioritizing its own passenger. We must consider that unlike human drivers this kind of artificial intelligence could analyze the situation and consequently decide how to react in a blink of an eye. These prerogatives deeply change the situation since we can’t blame humans for not having immediate reflexes but AVs manufacturers have the possibility to program them in advance. Furthermore they can’t just ignore these decisions because having the power to decide means that they are forced to face these critical situations.

This dilemma also leads into some inconsistencies: individuals may prefer and value the utilitarian method as the most correct, but when it’s time to buy a car that could protect people that you care about a little bit more, things could change. An interesting study was made in 2015 in this field: after showing images similar to Fig. 2 to different groups of people with some differences within each test regarding the number and the entity of victims (pedestrians or passengers) applicants were asked to evaluate their likelihood of buying an AV. Results show that subjects demonstrated a significantly lower wiliness to buy the AV if they were shown a situation where their relatives would be sacrificed to spare other lives. This test shows us that the utilitarian principle is seen as the most correct in moral terms but when it comes to consider the possibility of losing a loved person, applicants preferred the self-protective model. Furthermore we must consider that this dilemma is also transferred to manufacturers: a self-protective car would cause a public scandal while the utilitarian one could scare customers away.

Even if we ignore public preferences and we accept the utilitarian criteria as the best in general terms we should face more problems: this kind of cars will be able to identify the nature of objects close to them, recognizing animals and humans, but more surprisingly distinguishing each individual from their physical characteristics. Therefore being able to recognize the age and gender of both pedestrians and passengers allow us to decide which criteria should be programmed into the car in order to choose who should be protected at first. Who should be in charge to fix this criterion? Maybe an international institute is needed to establish standard to follow in critical situations. Following the common sense dictated by the public at large could be an alternative but opinions can be heterogeneous. Professor Iyad Rahwan is conducting the largest global ethics study ever made in this field: he created a Moral Machine website where anyone can test its own evaluation method passing thought different scenarios. This test has been a huge success collecting more than 4 million individual entries during the last year. First results were anything but clear-cut, except for 'simple' choices (like between hitting a child or hitting an adult) where the results were decisive, overwhelmingly favoring the protection of younger lives. However if we add more variables results tend to blur: in a more complex example involving also passengers more than 60% of applicants have chosen to sacrifice them to protect pedestrians. In general terms they stated that the more complex the scenario the less the decisions are one-sided.

We must remember that having the possibility to decide and still not taking any position makes us responsible anyway so we can’t just ignore this kind of situations. Furthermore as passengers give up any control on the vehicle the responsibility is totally transferred from the pilot to the car manufacturer so they definitely need to consider also extreme situations.

Conclusions

The path to a functional network of self-driving vehicles that can outclass our actual situation will probably be steeper than we can imagine. AVs will need to be introduced gradually while humans will try to agree on which ethical principles should be followed by this technology. The real paradox is that, despite we would be able to reduce impressively the number of road accidents, the unavoidable ones would cause more issues and scandals than the current crashes do.

In this brief research we saw that the most complicated side to resolve is not the most obvious one: in the near future humanity should study global solutions to both the social inconsistence of utilitarian vs. self-protective views and the moral dilemma of how to evaluate who has to be protected as a priority. As a matter of fact we are able to program these concepts into AVs but we don’t really know which side to choose and how to let them adapt to different situations. The key point is that we must firstly clear our position as professor Iyad Rahwan said:

“If we are to build machines that reflect our own values, then we need to understand those values more, and we need to quantify them and negotiate to agree which the important ones are.”