The rise of autonomous vehicles
Autonomous driving has moved from the realm of science fiction to a very real possibility during the past twenty years, largely due to rapid developments of radar technology and microprocessor capacity. Portable technology has sufficiently advanced to allow ultra-light hardware to make decisions based on self-improving algorithms, which means that developers stand a better chance of replicating the real-time decision-making of humans in autonomous cars.
Autonomous vehicles have been under development since the mid-twentieth century. Stanford engineering student, James Adams, first developed a fully autonomous lunar rover in 1961. Private corporations developed other autonomous vehicles and unmanned aircraft in the following years, but the technology remained a niche until Tesla launched Autopilot in its Model S sedan in 2015. Designed to navigate roads, exits, and stop-and-go traffic with minimal driver involvement, Autopilot marked the first time that high-level autonomous technology reached the hands of mass-market consumers.
Today’s autonomous vehicles use a combination of forward-facing cameras and radar systems that collect relevant information about road conditions for further processing. These systems can learn to detect, classify, and assign ‘weights’ to road obstacles, thereby giving the vehicle’s processors enough information to choose the path of least resistance. The values for these ‘weights’ correspond to how devastating an impact with a given object would be. Critically, each ‘weight’ is arbitrary, and the developers’ personal judgments play a profound role in how an autonomous vehicle will behave.
To keep pace with the rapid development of autonomy, the Society of Automotive Engineers (SAE International) developed six levels for autonomous vehicles, with level 0 indicating no automation at all, and level 5 indicating full autonomy, only requiring the entry of a destination. Tesla’s AutoPilot corresponds approximately to level 2 automation, while Audi is reportedly ready to see its level 3 autonomous car (an A8 luxury saloon enabled with the Traffic Jam Pilot), but is awaiting legal approval in many countries, including the UK and the USA. There are no level 4 or 5 autonomous cars available to the public, but the race for full autonomy is in force among car manufacturers such as Tesla and Audi, and technology companies such as Uber, Lyft, and Google’s Waymo.
Legal frameworks and other national initiatives
The speed at which autonomy has developed has made it challenging to regulate. In 2017, the US Congress started to debate the Safely Ensuring Lives Future Deployment and Research in Vehicle Evolution (SELF DRIVE) Act, a draft legislation aimed, among others, at transferring jurisdiction over autonomous vehicle testing from American states to the federal government. Furthermore, the SELF DRIVE Act would grant the National Highway Traffic Safety Administration (NHTSA) full control over autonomous vehicle design and construction, requiring the body to propose a series of autonomous safety guidelines within two years. Importantly, the legislation would mandate that autonomous vehicle manufacturers have a ‘privacy plan’ that details all potential uses for passenger data. Currently, the NHTSA can authorise 2500 autonomous vehicles for testing annually, but the SELF DRIVE Act raises the ceiling to 100 000 vehicles after four years. Still, the legislation has only been ratified by the House of Representatives; as of June 2018, the bill still needs the Senate’s support.
Despite the lack of consensus in the US Congress, states have introduced – or are working on introducing – their own legislations governing the testing and operation of autonomous vehicles. For example, in April 2018, California started processing applications for driverless vehicle testing permits up to level 4 of autonomy. In order to qualify for a permit, an autonomous vehicle must be able to resist cyberattacks, operate only within specific regions of a city, and come equipped with a form of two-way communication, as adjudicated by the California Department of Motor Vehicles (CADMV). In March 2018, Governor Doug Ducey issued an executive order in Arizona, outlining the conditions under which autonomous vehicles could be tested and operated on public roads. Some of these conditions include: the automated driving system behind the vehicle being in compliance with the applicable federal law and motor vehicle safety standards; if a failure of the automated driving system occurs, the vehicle will achieve a minimal risk condition; the vehicle complies with all applicable traffic and safety laws and regulations of Arizona, and it meets all applicable certificate, title registration, licensing, and insurance requirements.
In the European Union (EU), Germany has been a trailblaser in autonomous vehicle policy on account of its important automotive sector. As of 2017, Germany has a law in place that allows the testing and operation of autonomous vehicles on public roads, under certain conditions. Among them is the obligation of having a safety driver behind the wheel, to be able to take control of the car if needed. The legislation is to be revised in 2019, in light of technological developments. The current legislation does not legalise level 5 fully autonomous vehicles, but provides a deeper ecosystem for testing within German borders.
In the UK, the government announced in March 2018, that it had commissioned a review of driving laws 'to ensure that the UK remains one of the best places in the world to develop, test and drive self-driving vehicles'. The Law Commission of England and Wales and the Scottish Law Commission will conduct a three-year review into existing driving legislation, to identify 'any legal obstacles to the widespread introduction of self-driving vehicles and highlight the need for regulatory reforms'. Some of the questions to be examined as part of the review include: who the ‘driver’ or responsible person is; how to allocate civil and criminal responsibility where there is some shared control in a human-machine interface; the role of automated vehicles within public transport networks and emerging platforms for on-demand passenger transport, car sharing and new business models providing mobility as a service; whether there is a need for new criminal offences to deal with novel types of conduct and interference; and what the impact on other road users is, and how they can be protected from risk.
In May 2018, the European Commission presented a communication entitled ‘On the road to automated mobility: An EU strategy for the mobility of the future’, outlining a set of action points aimed at achieving the EU's ambition of becoming 'a world leader in the deployment of connected and automated mobility'. The Commission notes that current EU legislation is largely suitable to allow automated and connected vehicles to be put on the market, but that new regulatory changes would be needed to create a 'harmonised, complete and future-proof framework for automation'. Other areas of focus outlined in the communication include: (a) Allocating investments in technologies and infrastructure for automated mobility; (b) Ensuring an internal market for the safe take-up of automated mobility (by elaborating guidelines for a harmonised approach to automated vehicle safety assessments, for example); (c) Proposing new safety features for automated vehicles (by amending current regulations and directives on motor vehicle and road infrastructure safety); (d) Addressing liability issues, ensuring cybersecurity, data protection and data access; and (e) Exploring the implications of automated mobility on society and the economy (with a view to determining whether regulatory measures are needed to address the possible negative impacts).
The USA and Europe are hardly the only ones paying increasing attention to the ongoing developments in the field of autonomous vehicles. China is taking action to keep pace with other countries which support the development of self-driving vehicles, and, in April 2018, it introduced national guidelines for the the testing of such vehicles on public roads. One of the key requirements is that the test vehicles must always have a safety driver on board to be able to take control in unforeseen circumstances. These national rules come to complement guidelines already announced at the local level, in cities such as Beijing and Shanghai.
United Arab Emirates
In the United Arab Emirates, the Dubai Future Foundation and Dubai’s Roads and Transport Authority launched the Dubai Autonomous Transportation Strategy, in March 2018, with the overall objective of transforming 25% of the total transportation in Dubai to autonomous mode by 2030. The strategy will include, among others, the elaboration of specific legislation for the development of autonomous transportation, as well as the establishment of dedicated infrastructures for the new technologies.
In Japan, the testing of automated vehicles on public roads is done in line with guidelines adopted by the National Police Agency, which has recently created an expert panel to discuss and propose rules for autonomous vehicles, including in terms of penalties for accidents and traffic law violations involving level 3 and 4 autonomous vehicles. There are also plans for specific legislation to be proposed to the country’s legislative body by 2019.
In Australia, the government is planning to introduce a uniform regulatory approach towards autonomous cars, and national legislation in this regard is to be adopted by 2020. A policy paper published by the National Transport Commission in May 2018 outlines several recommendations for this legislative reform to ‘provide clarity about the situations when an automated driving system (ADS), rather than a human driver, may drive a vehicle; ensure there is a legal entity that can be held responsible for the operation of the automated driving system; establish any new legal obligations that may be required for users of automated vehicles’.
Liability, ethics and human rights implications
The novelty of autonomous technology stands to change our legal and social relationships to everyday transport. Importantly, without a driver behind the wheel, autonomous vehicles raise questions about responsibility and liability for road behaviour. If an autonomous vehicle were responsible for a fatal crash, for example, it is unclear how the legal system would assign negligence. Further, it remains unclear whether and how a company’s hardware and software developers, executives, and marketing employees would bear the blame.
A critical debate point over autonomy’s future lies in how the technology will approach life-or-death decisions. Envision a scenario where a child darts into the street ahead of a self-driving car. Facing the car is the decision either to swerve away from the child, risking passenger safety, or to continue on, potentially striking and injuring the child. Some commentators argue that the discussion should not focus on deciding ‘who dies’ in such cases, but rather on how companies can keep ‘no way out’ situations from happening at all. While edge cases will still happen (at which point there should be a clear protocol for assigning responsibility to the appropriate parties), companies should direct more resources towards developing the technology to the point where life-or-death scenarios happen very sparingly. The lack of clarity around how to simulate human judgment in all driving scenarios remains a serious concern for self-driving cars.
Autonomy may also pose a threat to privacy and personal data protection. Without a human driver to handle communication and customer service, the technology will have to rely on rider profiles to personalise the transportation experience. Web habits, social media accounts, location services, and the like will inform many of these rider profiles, so there will be every incentive to collect and store as much user data as possible. And the more data is collected to personalise the rider experience, the more opportunities for data misuse will arise. In addition, because autonomous vehicles do not need transparent glass, it is likely that paid advertising will surround customers during their rides, all of which will be personalised to the user’s interests.
Even further, the lack of a human driver opens autonomous vehicles to hacking, which poses a serious risk both to the rider’s and public safety.
Leading opinions on an ethical rollout for autonomous vehicles
Academia has been producing literature on the ethics of autonomy since the mid-twentieth century. Prominent among the thought leaders was author Isaac Asimov, whose 1942 ‘three laws of robotics’ aimed to set recommendations for human relationships with sentient technology. Asimov stated that above all else, creators of autonomous technology should specify who bears responsibility in the event that the technology goes awry before it is publicly deployed. Furthermore, any moral calculus used by an autonomous technology should be universally appropriate - that is, the ethical framework that the technology uses should be representative of rational human judgment. Finally, Asimov recommended that policymakers, technologists, and ethicists collaborate in predicting and controlling for a novel technology’s potential implications.
Furthermore, the literature suggests that codifying how best to approach ethical dilemmas in autonomy will increase consumer understanding of the risks associated with driverless cars. In 2017, the German Ethics Committee on Automated Driving published a report detailing 20 principles to ensure ethical programming in driverless cars. Some of the principles include: Automated and networked driving is ethically necessary if the systems cause fewer accidents than human drivers; in the event of danger, the protection of human life always has stop priority; in the case of unavoidable accidents, any qualification of people according to personal characteristics (age, sex, physical or mental constitution) is not permitted; in any driving situation, it is necessary to define and identify who is responsible for the driving task: the human being or the computer; and, drivers should have full control over what personal information is collected from their vehicles.
Academics have also simulated ‘worst-case scenario’ in order to imagine all of the ethical circumstances that a driverless vehicle could encounter. Paramount to this analysis is the ‘trolley problem,’ which describes a scenario in which a runaway trolley is careening toward five bystanders. A conductor has the option to let the trolley strike the five victims or to redirect it to another track, where another single bystander is trapped. The scenario cuts to the crux of the utilitarian argument, raising the question of whether it is morally permissible to sacrifice one life in the name of saving five, or vice versa.
Professor of philosophy, Nicholas Evans, from the University of Massachusetts at Lowell, used the trolley problem as a starting point for further testing. With help from engineers, Evans wrote multiple decision algorithms that draw from leading ethical frameworks in the academia, each of which prioritises disparate factors in making a moral decision. He then applied each of these algorithms to a repeated trolley problem scenario. Ultimately, Evans states that his goal is to display as many potential outcomes as possible, arming policymakers and the private sector with the tools necessary to select the appropriate framework(s).
Social and economic implications
There are important benefits to the autonomisation of vehicles. Fully autonomous vehicles will help to minimise the driver distraction that causes the majority of highway deaths. According to the US National Highway Traffic Safety Administration (NHTSA), 94% of highway deaths can be attributed to human error or poor driver decisions. Over 40 000 drivers died on American roads in 2017, meaning that a sophisticated autonomous driving system has massive humanitarian potential. Moreover, Americans spend 17 600 minutes behind the wheel of a car each year, on average, all of which autonomous vehicles could repurpose toward more productive ends.
From another perspective, autonomy is likely to democratise access to transportation, driving lower operating costs per mile than the current system of vehicle ownership. More urbanites will have access to safe and reliable transportation than ever before, driving widespread benefits to urban standards of living. Furthermore, autonomy is expected to make for drastically safer roads, helping to eliminate the driver error that causes the vast majority of fatal auto accidents.
Nonetheless, once public understanding of autonomous technology improves, pedestrians may begin to take advantage of the technology’s shortcomings. For example, if a pedestrian understands that an autonomous car will never elect to strike a human over an object, they might be tempted to cross streets at will, which would disrupt traffic flows.
Dangerous, too, are encounters between autonomous vehicles and human-operated vehicles, as shown during Uber’s first publicised autonomous accident in 2017. The fact that an autonomous vehicle will be more cautious than a human driver could also work to the technology’s disadvantage. Human drivers are accustomed to imperfect road behaviour from other drivers, meaning that a blindly obedient vehicle may pose a danger to itself and other drivers. Furthermore, autonomous vehicles are poorly equipped to interpret driving nuances, such headlight flashes, meaning that miscommunications may occur between drivers and self-driving cars.
Vehicle ownership and auto insurance
Autonomous driving could eliminate much of the private vehicle ownership in major urban centres. Now, where many city-dwellers deal with the inconvenience of storing and maintaining a car, on-demand autonomous vehicles would make it possible to ‘subscribe’ to a transportation ‘service’ instead of privately owning a car. The sunk costs associated with owning a car, then, could disappear, giving many urbanites more disposable income and further driving economic lifts in urban centers.
If autonomous cars do become a primary form of transportation in urban centres, carpooling may also become much more common. Regardless of an autonomous vehicle’s fuel source, the transportation sector will undoubtedly waste significantly less energy in urban centres, giving further benefit to public health and resource preservation.
Autonomy will also redesign current structures for auto insurance. Where insurance corporations now conduct extensive research about an individual driver in order to price policies, no analogous information exists for an autonomous vehicle. The insurance industry will need to redesign its current pricing structure to cover the decisions of a machine instead of a human, and a given fleet’s safety track record will likely play a major role in the calculus.
Last, but not least, driverless cars may jeopardise many sectors of the service-based economy. For example, as of 2017, 281 000 licensed taxi and private hire vehicles operated in the UK, with large sections of the lower middle class relying on taxi jobs for economic mobility. There are concerns that autonomy could eliminate many of these jobs, providing taxi companies with a cheaper, more efficient, and more reliable driver solution. Unless plans are put in place to retrain these displaced drivers, negative economic consequences are expected to ensue for the most vulnerable parts of the urban population.
However, there are also opposing views, which argue that the economic benefits from self-driving cars will surpass labour disruption concerns. A report released by the Securing America’s Future Energy organisation concluded that job dislocation and contribution to unemployment might not be as severe as commonly suspected, as new jobs and other economic benefits would compensate for any expected labour market disruption.
Ongoing research and impediments to autonomous vehicle development
Car manufacturers, tech companies and researchers are continuously working on further developing the technologies behind autonomous cars. In March 2018, for example, researchers at Stanford University presented a laser-based system that could allow driverless cars to notice unexpected obstacles before they come into view. California-based self-driving car company Drive.ai is planning to launch an on-demand self-driving car service soon in the city of Frisco, Texas, USA. There are many more such examples presented in the media on an ongoing basis.
Critical to the long-term success of autonomous driving are the millions of miles of testing in real-world driving scenarios. In the USA, de facto autonomous vehicle policy still falls into the hands of individual states, meaning that the leading corporations in the autonomy space have been able to test only in limited capacities. Thanks to its lax requirements for self-driving cars, Arizona was historically a testing ground for companies, but a recent autonomous Uber crash saw the company’s testing privileges revoked. However, as outlined above, more and more countries are exploring legal and regulatory frameworks that both support progress in the area of autonomous cars, and put in place safeguards for the testing and operation of these vehicles on public roads.
Furthermore, societal acceptance of autonomous vehicles tends to be low. According to a 2017 Gartner study, 55% of respondents did not feel comfortable riding in an autonomous vehicle at all, citing the potential for technological failures. The respondents fully acknowledged the technology’s potential benefits to transportation efficiency, but remained skeptical.
Finally, union activity continues to restrain autonomy’s pace. In 2017, the American Teamsters Union, 1.4 million members strong, fought to remove trucks from a bill designed to accelerate the autonomous vehicle rollout in the USA. Similar unions oppose passenger car automation in the name of job security. Private car ownership is deeply embedded in much of the Western world’s ethos, and autonomy will continue to suffer as long as lawmakers fail to address its relationship to the future of work.