Driverless cars: Who to kill?

When it comes to automotive technology, carmakers are in a race to launch the world’s first driverless car. However, before widespread adoption of this new technology can take place, carmakers are confronted with an ethical dilemma: How should carmakers program driverless cars to make moral decisions when faced with an unavoidable accident? Considering personal safety and public welfare, should it be programmed to minimise the loss of life, even if it means sacrificing the occupants?

Picture 1.jpg

Why driverless cars should prioritise the car occupants’ safety above others?

Since the advent of vehicles, human error has always been coined as the main culprit of traffic accidents, heavily outnumbering other causes such as faulty machinery and weather conditions. With the imminent launching of driverless cars, it is widely speculated that the number of road accidents would be significantly decreased. However, before putting ideas into practice, engineers are posed with an ethical design dilemma. Regardless of the moral decision made, the real life version of the Trolley problem will certainly have a knock-on effect on all road users, legislative bodies and carmakers. In this case, all stakeholders have a common goal of ensuring road safety.

Imagine you’re sitting in a driverless car, when suddenly a cyclist veers out of control and falls onto your path. How should your car react in such an unavoidable accident? One possible option is to integrate a moral algorithm that select the least destructive outcome in the event of unavoidable accident. A survey revealed that people are generally comfortable with this idea, but displayed reluctance to use such a car in fear of being sacrificed. However, some would argue that the moral algorithm should be based on priorities such as children, important public figures, and perhaps the car occupants. In contrast, some experts suggested the use of random outcome generators to prevent any discriminatory behaviour. Nevertheless, common sense dictates that matters of life and death should not be intentionally left to chance.

Using the utilitarian approach, incorporating a moral algorithm to minimise the collective ‘suffering’ seems morally permissible. Still, many might prefer driving a conventional car due to the possible risks of dying towards driverless car occupants. Consequently, this would lead to a comparatively larger number of traffic accidents owing to human error which increases the ‘suffering’ on a global statistical scale. Therefore, we can say that the utilitarian approach is inconclusive.

An alternative approach to the problem is using Kant’s theory which advocates on equal dignity and respect. Several experiments conducted implied that driverless cars are less likely to be responsible for causing accidents. The studies show that the events leading to accidents are likely to be caused by other parties such as unaware pedestrians jaywalking across the street. According to Kant’s theory, we have the right not to be harmed or injured unless we freely or knowingly do something to deserve punishment or choose to risk such injuries. As such, it can be argued that it is ethically impermissible for the car occupants to suffer the negative consequences due to negligence of others. This approach supports the notion that car occupants should receive a higher safety priority.

In retrospect, car occupants should be prioritised in the event of an unavoidable accident due to the mentioned reasons.

Why equal right for all road users should be applied to driverless car?

The emergence of driverless car might potentially compromise the safety of non-occupants. To date, there are no laws related to having driverless car on public roads within UK. Furthermore, the average number of non-occupant fatalities per year represents more than 51% of the total fatalities from all road user groups. Ideally, carmakers aim to increase their profits through higher production while complying with safety legislations. The desire of all road users will be minimal exposure to road accidents whereas the local government will seek to protect all road users without compromising the economic welfare of the stakeholders.  

           To ensure the safety of non-occupants, an institutional review board should be created so carmakers are compelled to provide potential accident scenarios and subsequent driverless cars’ responses. Safety legislations associated to the legal responsibilities of each stakeholder should be defined before the driverless cars are allowed to operate on public roads.  All stakeholders should be fully aware of the capabilities of the moral algorithm.

             According to utilitarian framework that emphasizes “the greatest happiness for the greatest number”, the quantifiable and comparable utility in an unavoidable crash is the number of fatalities resulting from the accident. With the statistical data shown above, it is sensible to consider the welfare of non-occupants can bring the greater happiness to the majority of all road users. Think about a road accident involving collision between a driverless car and a public transport such as an occupied bus, the human lives on the public transport could be determined as ‘expendable’ by the algorithm if the driverless car were to prioritise its own car occupants. This action opposes the principle of utility because it could result in higher fatality rates. With the designation of a review board, the board is able to evaluate the ethical judgement of these responses and subsequently ensure the rights of all road users are protected.

           It is suggested that driverless carmakers should reveal the capabilities of the moral algorithm. This is to ensure the stakeholders possess sufficient information to make autonomous rational decisions. If a non-occupant is harmed due to being unaware of  athe fact that the driverless cars are programmed to prioritize the safety of the car occupants, there would be no regulation enforced to protect them. This will effectively contradict with Kant’s reciprocity principle where humans should not be used as means to achieve certain goals. Failure to provide this information to the public implies that carmakers could exploit non-occupants via legal loopholes. Thus, they could avoid being legally liable for the accident.  With the information made available to the public, relevant stakeholders namely legislatures will be able to ensure the moral algorithm does not infringe the rights of non-occupants.

37: Sean Loo Wei Jie, Felicia Lee Pui Xuan, Lim Ken Hua, Ling Min Tan

Advertisements

9 thoughts on “Driverless cars: Who to kill?

  1. It’s definitely a difficult issue to solve, however I think the Trolley problem is too contrived to solve. One is looking for an ethical solution, however no such answer exists. Any choice or action taken will result in a loss of life. Therefore the argument takes the form of whether the action taken is ethically wrong or avoidable, rather than searching endlessly for an ethically right solution.

    On another note, I do find the general public’s view of the utilitarian approach rather ironic. The “I’ll risk my life, as long as it’s not mine” harps back to our basic human instincts of survival, but shows that even though we pride ourselves on being morally and ethically right, we are not willing to act when the time comes.

    Like

  2. Regarding this particular ethical dilemma, it seems that one of the main difficulties in reaching a conclusion lies in the pre-determined nature on a life-and-death issue. The survey is interesting in the sense that people show doubts about sacrificing ‘for the greater good’ when they themselves are the victims. However, perhaps one’s reaction in on a survey question and in a real-life setting can be different. As driverless cars are generally perceived as vehicles able to comply with road safety regulations; respondents thus possibly view that occupants are by definition not the culprit of the occurrence of accidents. Since the burden of preventing injuries now resides not with the driver but the prior programme design; people may not feel that they have the moral responsibility for a fault that is not caused by them. However, as numerous heroic actions in real life events tell us, the decision to risk one’s safety or even sacrifice is not beyond the realm of possibility. Rather, the most problematic subject on this matter appears to be the action to circumscribe the agency of one’s capacity to decide on a life-and-death issue.

    As demonstrated by the example involving a bus in the accident, the nature of the accident may directly affect the scope of ethical considerations if it is to address or be aligned with public expectations. The action of pre-determining the outcome of car accidents, in practice, seems to be a much less a clear-cut issue as would have been commonly assumed. I am not sure as to how technologies nowadays we can help us accurately predict and programme a driverless car to cope with every possible accident. However, it seems that the complexity of real-life events will inevitably raise difficult ethical questions. It also raises the question of how much responsibility the occupants should bear when accidents occur (perhaps such as the responsibility to have regular check-ups of their cars?). In this regard, the legal dimension of the safety of driverless cars definitely requires more attention from the government and the public.

    Like

  3. Regarding this particular ethical dilemma, it seems that one of the main difficulties in reaching a conclusion lies in the pre-determined nature on a life-and-death issue. The survey is interesting in the sense that people show doubts about sacrificing ‘for the greater good’ when they themselves are the victims. However, perhaps one’s reaction in on a survey question and in a real-life setting can be different.

    As driverless cars are generally perceived as vehicles able to comply with road safety regulations; respondents thus possibly view that occupants are by definition not the culprit of the occurrence of accidents. Since the burden of preventing injuries now resides not with the driver but the prior programme design; people may not feel that they have the moral responsibility for a fault that is not caused by them. However, as numerous heroic actions in real life events tell us, the decision to risk one’s safety or even sacrifice is not beyond the realm of possibility. Rather, the most problematic subject on this matter appears to be the action to circumscribe the agency of one’s capacity to decide on a life-and-death issue.

    As demonstrated by the example involving a bus in the accident, the nature of the accident may directly affect the scope of ethical considerations if it is to address or be aligned with public expectations. The action of pre-determining the outcome of car accidents, in practice, seems to be a much less a clear-cut issue as would have been commonly assumed. I am not sure as to how technologies nowadays we can help us accurately predict and programme a driverless car to cope with every possible accident. However, it seems that the complexity of real-life events will inevitably raise difficult ethical questions. It also raises the question of how much responsibility the occupants should bear when accidents occur (perhaps such as the responsibility to have regular check-ups of their cars?). In this regard, the legal dimension of the safety of driverless cars definitely requires more attention from the government and the public.

    Like

  4. Using an approach quite similar to how doctors use to prioritise who should be given life saving care first i.e. in case of a mother with a hindered pregnancy, mortally wounded soldiers at battlefield etc. could be relevant. Doctors do this to minimise the loss of life and take an objective approach, employing raw factual data and assertions which were made by people of similar position before them. The algorithm for driverless cars might have to take similar approach – i.e. it could be feasible that the car has to take into account the least destructive principle you described. The problem with Kant’s approach is that it takes way the idea of personal ownership for the driverless cars. The automated cars are still owned by the designated “driver” within the car. It would be impossible to put the blame on a machine or an algorithm, as it is well know that glitches are plentiful and unavoidable. As such, the article is right in contradicting Kant’s approach. Finally, a moral dilemma or not, in an economic sense, not only car makers pay a role, but also do insurance companies. The insurance companies are much likely to pay out to individuals to who are affected by driverless car crashes, as they can not be sure of the relevant party who is liable. It is more likely that they would prefer a morally least destructive algorithm over any other mentioned ones, as insurance risk principles are underpinned by that type of algorithms. As such, although driverless technology is advancing, there is a considerable risk that law is not developing with it, not only because of the ethical dilemmas, but also not asking the right questions.

    Like

  5. In my opinion, I do not think that the driverless carmakers can ever achieve a win-win situation if they are going to perceive this issue in both frameworks. There is no common ground to both the Utilitarian approach and Kant’s theory, in terms of the people that they seemingly aim to protect, i.e. Utilitarian – minimising overall damage regardless of who caused it, and Kant – protecting the people who have done no harm.
    If the carmakers were to apply these two approaches together, they have to take some extreme measures to keep external risks at bay. And the only way that I could think of is to have a lane exclusively for driverless cars, and the lanes will be tunnels to exclude any other forms of transport and also pedestrians. Even so, there may be other things that could have happened for example technical faults of the driverless cars, which can be life endangering, and that would contradict with Kant’s theory.
    If there is really a need to choose between the two, I personally would have gone with the utilitarian approach. Not only are we talking about physical injuries in an unavoidable accident, I would also like to take into account the mental trauma that will be associated with the accident. Knowing that the person would have survived if I were to share some of the impact of the accident, the whole incident will leave me in utter devastation. The Kant’s model, to me, seems a bit heartless. Some may argue that there might be pregnant women or young children in the car that are worth protecting, hence ensuring passenger’s safety is the ultimate priority. However, the other party involved in the car accident might also be pregnant women or young children. I understand the need for the driverless carmakers to ensure the safety of their passengers, and at the same time encourage sales of the cars by guaranteeing a 100% safety rate (which is rather unlikely), but to have the other party seriously injured or killed when you know that it is avoidable, just because they would like to show off their numbers (statistics), I do not think that it is ethically correct either.
    Ultimately, the choice that the carmakers make boils down to which group of passengers they are targeting at. Education and law enforcements have to come before the introduction of driverless cars as a preventative measure.
    *I think the essay is really well written and the ideas were supported very nicely.

    Like

  6. This is a decent write up on the safety issues that driverless cars are having. I would like to reenact the point where accidents normally happen due to human errors. Driverless cars will always aim to prevent the worst from happening but if non-believers try to sabotage this industry by endangering themselves or purposely causing an accident, accidents will definitely prone to happen. This is all due to one critical nature of humans, which is, wanting to be in control of everything. This trait is the main reason why humans have been so successful in creating new technology.

    One of the ways that can prevent the act of sabotaging from happening is to improve technology used to make driverless cars. 360° viewing and 24/7 full time operating cameras with cutting edge braking systems have to be installed on the car to prevent most of these from happening. A new set of rules also has to be practiced by all drivers on the road where driverless cars are up and running. A new set of procedures to claim from accidents that involve a driverless car have to be discussed thoroughly and easily accessible by everyone.

    With respect to the ethical theories discussed in the article, there is no one correct way. I believe different countries will practice different methods of handling accidents. It is all down to culture and beliefs. I do believe that driverless cars will reduce the number of road accidents immensely as the time taken for electronic systems nowadays to calculate the best scenario and react is multiple times faster than the time taken for humans to even realize an accident is about to happen. To even further reduce the number of casualties and fatalities, the driverless car can even contact emergency services before the accident actually happens.

    In contrary, opposition is actually good for this industry. The more the opposition, the higher the chance of creating the perfect driverless car.

    Like

  7. We all have our own perception of what is right or wrong. Quoting Oscar Wilde “Any preoccupation with ideas of what is right or wrong in conduct shows an arrested intellectual development”. There are basic fundamental things in a society that should be adhered to however sometimes what is right and what is wrong is dependent on many other factors too. However, jumping into the conclusion about these robot cars is not fit for the road could potentially stop an advancement that can provide a safer journey to all the road users.

    Currently there are many arguments about the driverless car ability to cope morally with complex real life event that comes with difficult ethical questions. The action of pre-determining the near-infinite number of potential situations that can result in an accident is putting extra pressure to the programmers as public questions the decision made.

    Car makers are pushing to launch their first driverless car, taking Google as an example, it has made clear that it is likely such a vehicle will not be street-ready for at least five years, and probably more. I have high expectations on the technology advancement that would made, where navigation and control technology will be perfected and the car would gain enough situational awareness in real life events.

    These car aims to reduce the human error factor on the roads. The public is questioning about the safety of having driverless car on the road, yet accidents is happening on the road every single day due to human errors. Noting that unresolved ethical issues have always been in play with automobiles. It is not wrong to be skeptical about the driverless car safety and ethical dimensions, it involved human life and you and I could potentially be the ones facing these situations. However, I would not rush into the conclusion that driverless car would not be able to handle complex real life situations. I am positive about the technology advancement and the level of road awareness in the next 5 years. This is definitely a giant task for the car makers and I am looking forward to see if these robot cars would be able to solve the ethical issues and prove that driverless can be a better choice in safety terms.

    Like

  8. An interesting article to read upon, nevertheless a new era to catch up with the new automotive technology. A way to think out of the box. However, it presents both advantages and disadvantages considering personal safety and public welfare, something to pour upon in which human being versus the programmable vehicle ? Are we ready for the huge change in term of safety legislations associated to the legal responsibilities of each of us, occupants of the driverless car is having equal right with the road users … etc.

    Like

  9. ‘the moral algorithm should be based on priorities such as children, important public figures’, I think this is an interesting idea, however in practice would the AI actually be able to make decisions as complex as this, especially in the split second before the collision? Perhaps this exposes a disconnect between the theoretical ethical problem and reality!

    Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s