The trolley problem in self-driving vehicles

Group 65

Autonomous vehicles or self-driving cars has shown various advantage comparing with traditional vehicles, especially in automated trucks industries. Tremendous investment and research have been done by automobile manufactures and AI companies. 

This article is considering an ethical thought experiment, which developed from the “trolley problem”. Imagine you are driving/sitting in an autonomous vehicle with your partner on a highway. Suddenly, some heavy object falls off the truck in front of you and you cannot stop in time. There are other vehicles, a motorcycle and another car by your sides which boxed you in.

You, or the self-driving system, needs to make a decision in a second. 

1.         Press the brake and stay in your lane. This will make the car hit the fallen object and may cause injury on you and your partner.

2.         Swerve to either left or right to hit other vehicles. This could minimize the damage on yourself but would hurt others.

Press the brake and stay in your lane.

Whatever actions taken by human here, they can be classified as reaction. But for the self-driving system, it becomes an ethical decision made by the programmers. 

For ethical reasons, autonomous driving systems should be designed to give priority to braking in such situations. While crashing a motorcycle into a side lane is a decision that can minimize damage, the choice is favorable for vehicle manufacturers, because this could guarantee their customers’ safety.

In contrast, the choice of slamming on the brakes may increase the driver’s own risk, but it can avoid involving irrelevant person into the accident. From the perspective of morality, this choice is obviously more suitable for the design of the automatic driving system.

This involves a moral theory: The Kantian ethic. The Kantian Theory is a kind of duty ethics, which can be used to determine whether an action is considered morally right or not. A core notion in Kantian Theory is autonomy, that is to say, the autopilot system should automatically comply with moral responsibility when responding. The system needs to make ethical decisions to get “good will”. The good will is a central notion in Kantian ethics. It means the actin should follow a categorical imperative. In this case, compliance with traffic regulations is the so-called categorical imperative. This is a rule, the starting point of all decisions of the automatic driving system should be to abide by traffic rules. As long as this rule is met, good will can be obtained. An emergency brake is certainly more legal than a sudden crossing of a lane into another vehicle. In any case, it is necessary to be responsible for hitting other vehicles, and the responsibility for braking is much smaller for the reason of the front car. In addition, according to Kant’s ethics, good will can be obtained by obeying the only principle, that is, obeying traffic laws. In other words, as long as the traffic rules are complied with, safety should be guaranteed. The motorcycles on other sides follow traffic rules and drive normally. If they abide by the sole criterion, but are still threatened by the safety of life, it is contrary to Kant’s ethics.

To sum up, braking is a more ethical option than hitting motorcycle in other lanes. The automatic driving system should be programmed to brake in such a situation.

Swerve to either left or right to hit other vehicles

Other ethics, such as utilitarianism theory, may result in the alternative decision on this thought experiment. A report from the University of Michigan Transportation Research Institute said the mortality rate of self-driving is lower than that of human drivers, but the proportion of car crashes is still likely to be higher than that of human drivers. But according to National Highway Traffic Safety Administration (NHTSA), they insist self-driving can help reduce 94% of car accidents. Anyway, Intelligent driving is the development trend of Automotive Industry.

The trolley problem in autonomous car has become the focus of discussion among general public and academia. In the most of car accidents, people will turn the steering wheel subconsciously to avoid from harm.  For an autonomous car, if the program choose to sacrifice drivers and passengers to protect others, the car company would lost some of potential customers, which may result in the reducing self-driving cars’ prevalence. The thing needs to identify is the technology can not recognize the humans’ gender, age and social status at the moment, so no argument will be shown here. According to the data above, if so, the car accident rate cannot be reduced as well. That means there will still have many accidents caused by human drivers per year, which can be avoid by autonomous cars.

In utilitarian terms, it’s about saving more lives. Hence, It’s not a bad choice to protect yourself in an unexpected situation.  

Furthermore, referring to the utilitarianism theory, the program should be designed for minimising the danger and minimising the adverse result. Due to the fact that the car carried two people while the motorcycle carried one person, crashing a motorcycle could save two people with hurting one person and this decision meets the utilitarianism theory standard which is producing the greatest happiness for the greatest number.

On the public traffic, all drivers, including the self-driving, should obey the local transport regulations and tried their best to reduce the irreparable losses. In this dilemma, braking could not save the couple’s life and it might lead to the rear-end collision. However, turning left or turning right could save the couple’s life. Even though the rider on the motorcycle may get hurt, the losses of the traffic accident have been reduced in maximum. In addition, turning the wheel gives the drivers who followed the truck longer vehicles’ distance and reaction time to avoid the collision.

The decision given by the virtue ethics, care ethics and Kantism’s theory is opposite to the crashing. After the braking, the strangers could not criticize the programed action but it might lead to severer result. However, turning left or turning right could minimize the adverse result and save more lives.

Initial Decision

Engineers and system programmers suggested that a random decision in similar situations might be the best solution at this moment.

26 thoughts on “The trolley problem in self-driving vehicles

  1. What’s interesting from reading your article is that the autonomous system needs to know the number of potential persons on either side. A motorbike can carry two people, whereas a car could have one driver only.
    Leaving ethics to one side for the moment, swerving to hit the car might make more sense since the car also has bodywork protecting the occupant/s, whereas the bike has less protection.
    I enjoyed the reasoning behind “Press the brake…”

  2. Anyway, hitting people is not good. Can personal security be addressed in other ways? The use of airbags and so on, should be able to minimize their own safety.

  3. When it comes to autonomous driving, you have to think about the whole route. In this case, although the brake is morally impeccable, it is still troublesome for the whole traffic system. People in the car will be seriously hurt, and it will be difficult to move the car out of the main road in the future, which will continuously increase the risk of accident.

  4. In your scenario, I’d choose to press the brake. From my personal driving experience, swerving the wheel would always result in further damage to the oncoming traffic.

  5. This could be a concern for AI industry. AI also need to make a choice among hurting more relevant people, or fewer irrelevant people, or reduce damage as much as possible. Hope you can suggest one of them.

  6. This is a very controversial issue. I think whether we need to turn both sides should be based on the actual situation at that time to make a judgment. But if there are few passengers on the bus and the road is not crowded, then turn both ways. But if there are a lot of people, in order to save the lives of more people, they should go straight. The design firm’s payment should be a secondary consideration.

  7. Considering moral reasons, I choose the brake. Although the option of braking may cause the car company to lose some potential customers, it avoids involving other innocent people in the accident, and vehicles that obey normal traffic rules should not be threatened with life safety.

  8. Turning left or right to hit other vehicles. I choose to save more lives by considering utilitarianism, and the choice will not lose potential customers, and it will promote the popularization of self-driving cars.

  9. In this dilemma, braking may not be an appropriate choice because braking may lead to more serious consequences, such as rear-end collisions. Turning left or right to crash into another vehicle may result in a motorcyclist being injured, but it saves the couple in the car and minimize the damage.

  10. In terms of public transportation, all drivers, including autonomous driving, should abide by local traffic rules to minimize irreparable losses. In this dilemma, braking would not save the couple’s life and could lead to a rear-end collision. However, turning left or right can save the couple’s life. Although motorcyclists may be injured, the damage from traffic accidents has been minimized. In addition, drivers following behind have longer driving distances and reaction times to avoid collisions.

  11. Considering the inertia of the weight, the object is bound to crash, and considering the minimization of the damage, I think the driverless car should turn in time. After all, the motorcycle riding beside the car is a small probability event, and the slogan for the motorcycle to give way should also be posted on the driverless car

  12. First of all, this article is very well written, and the discussion of the problem is more comprehensive. However, in my opinion, this is not the fault of anyone in this case. Everyone has the right to choose to protect themselves. For unmanned vehicles, the safety of passengers should be the first priority, both from a commercial and design perspective. No one wants to take a machine that cannot guarantee their safety.

  13. This is a question worth considering. But from another angle, if we can make self-driving cars popular, then in this case all cars choose to protect the owners, can they minimize losses? Since this question is difficult to give an answer, car manufacturers should try to avoid this happening.

  14. It is a difficult question to answer. maybe I can’t consider more things when the accident happen. for me, my first reaction was to press the brakes. because I just want to stop the car. I don’t have enough time to consider the accident will hurt me or others.

  15. Autopilot is a good development direction. However, after all, robots can’t be compared with people. In the face of unexpected incidents, there will be some shortcomings, especially traffic problems, most of which are caused by unexpected events.

  16. As technology developing Autopilot will become a trend in the future. Security is indeed a problem to be considered. The situation you are proposing is very valuable. In this case, what kind of decision to make is indeed a moral issue? I prefer the first choices, it saves more lives to the maximum, but it also hurt the innocent people. It is a hard decision when you met this problem.

  17. This is a classic moral dilemma. However, I think, as an engineer, we should design a program to avoid this kind of thing. For example, by reading the GPS data to predict which road is the safest to take, or controlling the speed to prevent the occurrence of injury.

  18. The dilemma was provided from the article is interesting. In my opinion, the automatic car should choose the automatic brake, because the car’s airbag can guarantee the safety of waiting, if entering the lane on both sides, it may cause more damage.

  19. The first duty of an autonomous car is to protect the safety of its driver. I think there are some preconditions that should be established before discussing the question of what kinds of risk avoidance should be adopted by autonomous vehicles. First, the operation of the automatic car is absolutely in line with the traffic rules. Second, the safety system of the automatic car is good enough. In this way, when faced with an emergency, should choose to break in situ, so as to ensure the safety of the opportunity, but also to ensure the safety of the property.

  20. In my opinion, autonomous cars should turn to the lane on both sides for the following reasons. First, for the whole traffic system, such a treatment can avoid greater damages, such as rear-end collision. However, that can only happen if the car detects whether road conditions on either side of the body allow it to turn into. Second, it can directly avoid a traffic accident.

  21. I think there are other options, like adding a small rocket propulsion device to the vehicle’s site. In this way, once an accident occurs, the injury can be avoided by flying.

  22. Autonomous system is an interesting topic as it is developing rapidly in these years.
    Actually there might happen some accident situations to a car while it is driving normally. To try to avoid these accidents, autonomous system need to have following conditions:
    1. Networking with traffic command.
    2. Keeping collecting cars’ data which driving nearby you through satellite positioning(every body has phones, so can positioning by phone).
    3. Sufficient and sensitive sensors.

  23. Autonomous driving is undoubtedly a trend in the automobile industry. The article is innovative in discussing the ethical and moral issues of autonomous driving technology from the perspective of traffic accident, the worst outcome. For a car accident that has already happened, how can the program minimize the loss is a question with no answer and it’s ethically controversial. The best solution is to avoid accidents.

  24. Making random decisions to increase uncertainty is to reduce the car company’s condemnation of condemnation after similar incidents. In reality, we may need to consider more circumstances, the environment, the weather, and the left and right sides. The number of people driving, the lanes followed by the car, my personal recommendation is to keep random decisions and increase temporary control to use our first reaction to deal with similar situations. A better approach is to use algorithms to avoid similar situations.

Leave a Reply