The Modern Trolley Problem: Ethical and Economically-Sound Liability Schemes for Autonomous Vehicles.

Author:Moolayil, Amar Kumar

    The automobile fundamentally altered the way people travel by replacing the horse with the engine; "autonomous vehicles" are in a position to make the same transformation by replacing human drivers with computer algorithms and sensors. Experts and risk analysts from the private and public sectors believe the removal of human drivers will eliminate almost all automobileaccidents (1) However, even if all human drivers were removed from the road, there would still be a substantial number of automobile accidents, some of which would be caused by autonomous vehicles. These accidents raise a difficult question which automobile manufacturers, insurance companies, programmers, and the legal system need to answer before autonomous vehicles dominate the road, namely who is ultimately responsible for such accidents.

    This note will contemplate various solutions to the problem of accidents caused by autonomous vehicles. The first section of this paper will provide a brief survey of the current landscape of autonomous vehicles: how the industry and government currently categorize various computer systems which control the vehicle; recent accidents involving autonomous vehicles; as well as the government's current position with respect to autonomous vehicles. The second section of this note will examine: 1. how the vehicle's programming is designed to behave in the event of an accident; 2. whether it opts to protect the passengers of the vehicle or whether it seeks to limit amount of damage caused by the accident, even if it means sacrificing the health and safety of the passengers; 3.who will ultimately be liable in the event of an accident, and; 4. whether the individual driver is liable for the accident or if the auto-manufacturer and its program designers should be responsible. These four categories will be combined to create four "liability schemes:" protect the individual and individually liable (the current system); minimize losses and producer liability (the corporation's agent approach); minimize losses and individual liability (the analytic a posteriori approach); and protect the individual and producer liability (the fiduciary agent approach). (2) The final section of this note will analyze each of the four liability schemes based on the ethical, legal, and financial consequences of each system for both the individual drivers and autonomous vehicle manufacturers. Finally, based on the insights gained from the analysis, this paper will make a recommendation on how the industry and legal system should adapt to the inevitable introduction of autonomous vehicles on the road from a liability perspective, and conclude by recommending one or two of the best liability systems.


    The technology behind autonomous vehicles has progressed quickly over the past year, and numerous automobile manufacturers have incorporated some or all of these technological innovations into their own vehicles. (3) The technological innovations of the past year have also allowed companies like Google and Tesla to begin "beta-testing" autonomous and semi-autonomous vehicles on the road, sometimes with controversial results, including the first recorded death caused by an error in the vehicle's auto-pilot system. (4) In addition to technological innovations, autonomous vehicle producers have received some help from the US government in the form of new rules and regulations issued earlier in 2016, which should pave the way for more autonomous vehicles on the road in the near future.


      Autonomous vehicles are currently on the road today, although not in ways a person may typically expect; features that many consider standard on many modern vehicles such as lane departure systems or adaptive headlights are considered by the National Highway Transportation Safety Administration (NHTSA) to make a vehicle "semi-autonomous. (5) " Currently the NHTSA uses a six-level classification system for identifying autonomous vehicles, which it adopted from SAE International in September of this year. (6) SAE International's system--ranging from Level 0, where all of the vehicle's functions are controlled by the driver, up through Level 5, a fully autonomous vehicle where the vehicle occupant has no control over how the car navigates the roadway--will be used by the NHTSA until it develops its own system (7). In addition to the five levels, the NHTSA categorizes vehicle on whether or not the "human operator or the automated system is primarily responsible for monitoring the driving environment;" human operators are the primary operators of vehicles that fall between Levels 0 through 2, automated systems are the operators of the vehicles for Levels 3 through 5. (8) Additionally, the federal government considers all vehicles that fall within Levels 3 through 5 as "Highly Automated Vehicles" (HAV). (9)

      Until very recently, all manufactured vehicles fell within Level 0, where the human operator performs "all aspects of the dynamic driving task, even when enhanced by warning or intervention systems." (10) At Level 1, the human driver is expected to control the majority of the dynamic driving task, but may be augmented by some aspects of an automated system, such as stability control or dynamic braking. (11) Vehicles that have a computer system built into them, including many vehicles built within the last five years, would fall within this classification. (12) At Level 2, the autonomous system is capable of controlling multiple tasks--steering, acceleration/deceleration, parking, etc.--using information it gathers from the driving environment. However, the human driver is still responsible for whatever remaining dynamic tasks the system is unable to perform at the time. (13) Vehicles that fall into this level have only recently entered the market. For example, Tesla released an "Autopilot" feature in October of 2016, which allow the vehicle's internal systems to operate the vehicle in place of the driver when activated. (14) As of October of 2016, the Mercedes-Benz S65 AMG, BMW 750i, Tesla Model S, and the Infiniti Q50S--among a few others--also qualify as Level 2, semi-autonomous vehicles currently available for sale. (15)

      Level 3 vehicles are the most basic form of the NHSTA's HAV designation. In addition to controlling the majority of the dynamic driving task, the vehicle is also capable of monitoring the driving environment and adjust to any changes. There is "an expectation that the human driver will respond appropriately to a request [from the vehicle] to intervene," and avoid a situation which the computer is unable to handle. (16) Currently, no auto-manufacturer offers a Level 3 vehicle for sale, but Google's current driverless vehicle would classify as a Level 3 autonomous vehicle; the "Google Car" uses adaptive sensors to determine its position relative to the road and other objects around it and the internal computer software "chooses a safe speed and trajectory for the car." (17) Level 4 vehicles have all the features of their Level 3 counterparts, but crucially do not require human intervention in emergency situations. In the event of a potential accident, this means that the human driver has the option to take control over the vehicle, but the vehicle does not require him to do so. (18) No vehicle has yet to reach this level, but Tesla CEO Elon Musk believes that his company will be able to produce vehicles at this level as early as 2018. (19) Finally, at Level 5, the human driver is completely eliminated from all functions of the dynamic driving task, as the vehicle can handle all driving situations that a human driver could, including emergencies. (20) The critical difference between Level 4 and Level 5 vehicles is that the human passengers have no ability to control any of the cars dynamic driving functions because the steering wheel and pedals have been removed. (21)

      As stated earlier in this note, the classification system currently used by the NTHSA is subject to change when the agency finally develops their own independent system, however it is not unreasonable to imagine that the categorization scheme they eventually adopt will bear a striking resemblance to the current SAE International Guidelines. This note will mostly focus on the liability for autonomous vehicles that fall within the Level 4 and Level 5 range, as these types of vehicles have the greatest potential to upset the current legal system.


      Advocates and skeptics of autonomous vehicles have received plenty of evidence for their positions based on beta-testing performed in 2016. Both Google and Tesla have tested some form of their autonomous vehicles in the last year; Google through its "Google Car" and Tesla through the activation of its "Autopilot" software. While industry insiders and the public have generally reacted positively to such developments, the testing of these new vehicles not been without issue. Autonomous vehicles have been involved in a few accidents in 2016, including a fatal accident in May that has been attributed to a failure of the vehicle's sensors to properly identify another vehicle on the road. (22)

      Google has been testing some form of autonomous vehicle since 2009, often retrofitting existing vehicles with their technology. (23) Currently, Google operates a fleet of 58 vehicles--"24 Lexus RX450h SUVs and 34 prototype vehicles"--all of which are being tested in various locations around the United States, including: Austin, Texas; Phoenix, Arizona; Mountain View, California; and Kirkland, Oregon. (24) As of August, 2016, Google's vehicles have driven over 1.9 million miles in "Autonomous Mode," with an average of 20,000 to 25,000 miles per week. (25) Google also introduced a vehicle close to the Level 5 SAE classification without pedals or a steering wheel...

To continue reading