• Home
• About
• Login
• Register
• Search
• Current
• Archives
• Announcements
• Pre-2013 Index

Home > Vol 20, No 3 (2014) > Bradshaw-Martin

Autonomous or 'driverless' cars and disability: a legal and ethical analysis.

Heather Bradshaw-Martin [1] and Catherine Easton [2]

Cite as: Bradshaw-Martin H., & Easton C., "Autonomous or 'driverless' cars and disability: a legal and ethical analysis", (2014) 20(3) Web JCLI.

Abstract

'Driverless' or autonomous cars have been in existence for decades but, more recently, policy moves have been made towards supporting the expansion of their use on public roads. Both the EU and UK have announced research into the technology alongside reviews of relevant law and policy. This article focuses on one very specific aspect of the debate; the use of autonomous cars to support the independent life of disabled people. While focusing on this one particular issue, it raises important ethical and legal questions around both the wording of key definitions and the fundamental rationale behind the employment of these systems. It is argued that the current definition of the concept of 'driver' is now obsolete in the face of technological progress.

1. Introduction

The technology relating to autonomous cars has now developed to a point at which many major car manufacturers are trialling their own systems [3] and the landscape has begun to shift, with commitments announced to commercialise these technologies. [4] Autonomous technologies developed by Google have already covered over 500,000 miles on public roads [5] and, with a number of States in the United States passing laws to regulate their use [6] and the European Union funding experiments into transport systems to support 'driverless' cars, [7] the day is looming when this technology becomes commercially available. In the UK, the systems are being trialled in Milton Keynes with the aim of implementing 100 automated vehicles on the town's roads by 2017. [8] Furthermore, in July 2014 it was announced that the UK Government intends to encourage the use of 'driverless' cars through a £10 million investment in research and a wide-scale review of the law and policy relating to regulating the roads. [9] There is, therefore, a pressing need for the early identification of key legal and ethical issues to ensure that they are able to shape the debate from the outset.

While the potential implementation of cars without human drivers raises numerous legal and ethical issues, this paper seeks to focus upon the technology's ability to support the independent lives of disabled people. In essence, autonomous cars could allow an individual with impairments to travel without the need for any additional human agency. However, the achievement of this aim may be hindered by pre-existing approaches to the legal and ethical definition of driver and driving. The key assumption in question is that the driver of a car, a civilian road vehicle, must be a person, that is to say, a human being. This assumption in enshrined in the Vienna Convention on Road Traffic of 1968, as amended in 1993. [10] This document, alongside the earlier Geneva Convention, [11] governs the design, regulation and use of road vehicles in most countries today, with a few key exceptions, such as China. At the heart of this premise are certain ethical assumptions which seem to have remained largely unquestioned [12] for the last fifty years.

In order to analyse the impact of legal definitions on the ability of this technology to enhance the lives of disabled people, an overview will first be given of research relating to the importance of travel in leading an independent life. This will then be built into an analysis of key theories relating to disability and a discussion of legal and political issues relating to self-driving cars. These will be drawn together to focus on the key question of whether or not the software and other systems controlling so-called driverless vehicles ought to be classified legally, and perhaps ethically, as the driver of such vehicles. There will be a strong argument that the assumption that such vehicles are driverless is rapidly becoming unworkably anachronistic, with unethical consequences.

2. Disabled people and Autonomous Cars: Framing the Issue

The potential for self-driving cars to provide mobility benefits for disabled people has been indicated in research undertaken during the technology's development. [13] This can be drawn into work on the importance of mobility in supporting independence and the ability to function fully in society. [14]

During research into human enhancement technology and disability, [15] a number of in-depth qualitative interviews with persons with various physical disabilities were conducted. Although not the initial focus of the research, inequalities in access to independent mobility were frequently mentioned by participants. One interview was with a young, degree educated, congenitally partially sighted man with young children and elderly parents. The following exchange summarized an unexpected and remarkably consistent emerging theme:

Interviewer: 'Is there anything if, if you had access to lots of money and you had, technologies that could give you …more intelligence, more strength, different abilities, new abilities…what would you want to do with that?'

Interviewee: 'It's just…ah you know… just…just be able to see in my case. But even that, it doesn't bother me now, like, you know because the only thing I can't do is, like drive a car.' [16]

On further examination of the issue, it is not just the experience of driving the interviewee misses as friends have offered him that opportunity on beaches and in empty car parks. Indeed, there are officially organised motorsport events for partially sighted and blind people with sighted guides. The nature of this experience is, however, problematic in its brevity:

'I'm like, no, I don't want to. I can't see the point... for 10 minutes, because you're not going to do it again. I don't want to get the bug, and end up joyriding or something. Like, oh, give him a car to play with and take it off him then. So that's the only thing.' [17]

When asked about attitudes to transport, the interviewee strongly focused upon the utility of personal mobility:

'My parents ...and both of them can't walk far or do much. I gave them a stair lift and things like that now.'…'if I was living next door to them I couldn't do certain things. The biggest thing they need help with is going shopping. My cousin takes them in the car and brings them home. But obviously I... I... I couldn't do that if I lived there. Because I don't drive like, you know? Can't go myself like. I can't say to them "Do you want a lift to town off [my wife's] father?"' [18]

This response highlights that access to independent personal mobility matters for human well-being today. Lack of adequate independent transport excludes people from employment, social life, education and access to medical care. Such exclusion has significant negative social, health and economic impacts, especially for older people. [19] The only thing the interviewee really cannot do is obtain a driving licence. However, catching the bus is nearly impossible, especially when leaving the city to come home:

'You've really got no chance, with the buses. In [local city] you've got four buses coming from the same stop. You've the thing on top now, that tells you: "the [northern town] is coming in 12 minutes" but... doesn't tell you when it's come in! Doesn't tell you it's here now! It's coming in 12 minutes…I remember once…I made this big sign, [with the number of the required bus on it] and the bus still didn't stop!.. [In fact the participant was reprimanded for his sign by one bus driver.] Ah, doesn't matter. Buses are just a law unto themselves like.' [20]

What became clear in the research was the importance for human well-being of access to independent personal mobility for people today; almost every participant in the research, which was not originally about transport, complained about buses. There is a major ethical problem in excluding people from such an important aspect of the way of life the rest of us take for granted today. Moreover, this exclusion has significant negative social, health and economic impacts, especially for the growing population of older people. [21] Public transport has singularly failed to address this injustice so far. A striking example can be found in the decision handed down on the 8th December 2014 in the case of FirstGroup PLC v Paulley. [22] This, at a basic level, revolved around an analysis of the interests of bus passengers who use wheelchairs and of those who do not. The discussion focused on the bus company's policy in relation to the use of a dedicated wheelchair space by passengers not in wheelchairs; in the current case, those with pushchairs. In a unanimous decision, the Court of Appeal overturned the finding at first instance that a policy requesting rather than enforcing non-wheelchair users to move from a space dedicated for wheelchair users was a "policy, criterion or practice" [23] which placed disabled people at a substantial disadvantage. Reasons given for the decision included the removal of a bus driver's discretion, the need for a bus driver to adjudicate potential competing claims and the practical problems inherent in enforcing a strict requirement. Surprisingly, Lord Justice Lewison, when discussing the potential impact of passengers being asked to leave a bus to make way for a wheelchair user, found that whether or not the wheelchair user was "in a hurry" could be relevant, rather than approaching the situation from a baseline of equality of access. [24] In an acceptance of the role of the court to apply the laws created by the legislature, Lady Justice Arden states:

'It follows from the judgments of this court that the proper remedy for wheelchair users is to ask Parliament to strengthen the powers of bus drivers so that they could, for instance, require people to vacate the wheelchair space, or create new duties on other passengers, or to campaign for a different design of buses.' [25]

and accepted:

'I do not underestimate the difficulties of travel for wheelchair users or their frustration at the pace of change. It is obvious that, as one wheelchair user has said, for them the world was not built with a ramp'. [26]

Notwithstanding this admission, this case demonstrates that a strict application of anti-discrimination law can lead to substantive barriers to access and there is a potentially wide-scale need for legislative change. On a wider, yet related issue, transport schemes put in place for disabled people can be a cause of much friction. [27] Basically, these always require two people to travel when only one wishes to undertake the actual journey. The cost of paying a second human to accompany the traveller is prohibitive. Given the hardships caused by inaccessible public transport, there is the potential for autonomous driving systems technology to be employed to address this on-going, divisive aspect of society.

3. Technological development and disability: Law and policy

Increasingly, barrier free technology is essential to the achievement of an independent life. The Internet has provided unprecedented opportunities for communication, activism and awareness-building. At the same time, both the public and commercial sectors are turning to technology as a cheap way of providing services and interacting with citizens and consumers. Without, for example, access to the Internet people are excluded from Government information and certain services. [28] They are also denied discounts, offers and up-to-date travel information. In this way, the emancipatory benefits of technology can be lost because the diverse needs of end users are not strategically addressed. [29] The nature of these needs, and of disability, is multi-faceted and can also be related to both permanent and temporary states. It is, therefore, important to consider the interests of actual populations, such as the ageing, when evaluating the legal and ethical issues surrounding the implementation of technologies such as autonomous cars.

In the development of anti-discrimination provisions, disability has often been one of the last areas to be addressed. In the UK, for example, wide ranging statutory provisions addressed race and gender discrimination two decades before a statute focusing on disability was introduced. [30] Similarly, the EU has its general gender and race-specific Directives [31] but as yet does not have the equivalent in relation to disability. In the past this has been attributed to factors such as the domination of charity and medical models of disability and the difficulty in defining such a multi-faceted state. [32] This has been addressed in certain jurisdictions with the use of the concept of reasonable, sometimes anticipatory, adjustments to address exclusion. [33] While the social model of disability has had significant influence on law and policy, certain technological advancements, such as the Internet, have been designed without a focus on access. In this way the experience of disabled people is often seen as a costly afterthought with retrospective changes begrudgingly made to accommodate those who deviate from an accepted norm. [34] It is important, therefore, that the development of technology relating to autonomous cars is developed with inclusion and access at its heart rather than as an afterthought. It would be hoped that in relation to these systems this would occur due to a moral impetus pushing towards inclusion and participation rather than a realisation that disabled people's use of the technology is a selling point which could be used to foster support in the face of safety concerns.

Autonomous cars present an opportunity to support the independence of a large and diverse sector of society. Improvements to road safety have been proposed but are contentious. [35] If the ethical arguments are not seen as definitive, at a more commercial level, the purchasing power of disabled people could be influential. In the EU this is estimated at over 300 billion euro [36] and in the UK the spending power of those over 65 has increased at an unprecedented level over the last decade, [37] autonomous cars as a commercial concern have the opportunity to tap into these markets.

4. Cars: A Disabling Transport Technology?

The premise that cars are excluded from the requirement to be designed for all is not a necessary one, it is, indeed, arbitrary, and as such creates an injustice. [38] It is arbitrary because cars do not have to be designed such that sight, or even the judgement and memory attacked by conditions such as Alzheimers, are necessary for their use. One of the first cars built in the USA at the turn of the last century was in fact built by a blind person, Ralph Teetor, and, according to his co-conspirator, driven by him using a guide wire to assist in navigation:

'How much did Ralph actually drive this car? Probably more than his parents knew or approved of. As has been said, he had learned to walk or run all over town by himself, determining where he was by the echo of his footsteps from the sides of buildings, trees or other landmarks. When there was no traffic he could drive through familiar territory in a similar way with the car. Dan did mention to his son Jack that the boys constructed an overhead line which was used to direct the car along a path, enabling Ralph to drive the car alone.' [39]

Ralph Teetor was an exceptional man, an excellent engineer and a successful company director. Later in life he invented and patented one of the first of the technologies which eroded the human's role in driving: cruise control. The design took him more than a decade to perfect, and his own concerns about drivers falling asleep without constant physical feedback were eventually overridden by customer demand and an excellent safety record. Contrary to modern claims [40] there was a furore in Congress about the new device at the time not dissimilar from that today about Automated Cruise Control and other aids which change the driver's role to passive monitoring rather than active control.

Today cruise control has become automated cruise control, and lane departure warning systems, anti-lock brakes, stability control systems, and navigational aids have become mature technologies. There is considerable evidence that car manufacturers had the technological ability to connect all these to create autonomous cars in the late 1990s. [41] These sources claim that legal concerns held back autonomous car development. There are two areas of legal concern: legality [42] and liability [43]. Garza claims that in the US existing liability law is sufficient [44] but this presumes legality.

A strong argument is put forward by Garza that autonomous cars, by cutting out the unpredictable element of the negligent or susceptible human, could, in fact, lead to safer driving environments. This may be attractive to insurance providers but this technology will not become a reality until it is accepted by the public as unproblematic and a strong legal framework in relation to liability will be crucial in achieving this. It has been suggested [45] that the impetus for legislative initiative will not arise until the larger car manufacturers grasp the commercial potential for the technology and sway political opinion. Lobbyists for Google and a number of car manufacturers have already made an impact in the USA at the State Assembly level with three States passing legislation to support the implementation of autonomous vehicles. [46] While public faith in the technology is difficult to gauge, it will, it is argued, be shaped by its prevalence and the regulatory environment surrounding it. When premiering his think tank-developed concept car at the Geneva car show, Frank Rinderknecht admitted that public acceptance would take a 'leap of faith' [47] and highlighted the need for tailored safety statistics to move opinion in its favour. The achievement of these specific standards could prove problematic. According to an analysis based on Poisson distribution, an accepted probability theory, there would be a need for an autonomous car to travel 725,000 miles without incident to be 99% sure that it was safer than a car with a human driver. [48]

Steve Mahan is the head of the Santa Clara Valley Blind Center, he has lost 95% of his sight and was recently given the opportunity to direct a self-driving car. [49] He is very aware of the powerful link between car driving and independence but does not believe that there will be wide-scale adoption of the technology for disabled drivers soon. Mahan believes that there is a need for society to become comfortable with the increasingly automated functions of cars before accepting disabled drivers. In a BBC report focusing on disability and technology, the following question is posed: 'So in the future we could have empty cars run by a computer, or cars with a disabled driver unable to over-ride the computer. Is there a difference, and which would you prefer ?' [50] This problematic framing of the key issue moves away from an emphasis on the nature of the technology and poses two equally likely scenarios. These should, from a practical perspective, not be treated differently as, in essence, technology would be in control in both cases. Clearly, an autonomous vehicle with no human passengers would ethically be expected to give way to and sacrifice itself if ever it encountered a situation where it endangered the safety of other road users. However, an autonomous vehicle with a human passenger, any human passenger, disabled or not, must be programmed to behave in a way which took account of those human passengers' interests as moral agents with equal moral status to other human moral agents.

For a given moral agent travelling on roads populated by many hazards and other moving vehicles, it would be rationally preferable if he or she did not have to assign the occupants of those other vehicles equal moral status in cases of conflict. At present such choices are rarely an issue as we can almost always assume that any oncoming vehicle does indeed contain other moral agents and thus must be treated according to the Geneva Convention's requirement that:

'When approaching other road users, they shall take such precautions as may be required for the safety of the latter'. [51]

In a future with autonomous empty vehicles commonly on the roads such circumstances and decisions may become more complex as not all vehicles will have equal moral value in this sense. A vehicle which contains humans but over which those humans have no control, perhaps because they are sleeping or engrossed in email or other activities, or are perhaps for other reasons physically unable to retake control, does place additional burdens on the drivers, human and non-human, of other vehicles, which must be able to 'take such precautions as may be required for the safety of the latter' without relying on the human agency and reactions of the other vehicle's controller. Disabled vehicle directors are not in any way dissimilar from sleeping or otherwise distracted vehicle directors in their inability to respond in an emergency.

The framing of this issue in this way focuses upon disability and gets to the heart of the question about the existence or not of a non-human driver. If such autonomous technologies are to be legitimated then we propose that legal approval of driverless cars for disabled drivers would need to be built upon a prior public and legal acceptance of computer-controlled empty cars.

5. Supporting an independent life

The truly emancipatory aspects of self-driving cars can only be achieved with a full and frank debate about the technology's ability to support disabled people's ability to live the independent life of which they are capable. This can be analysed through a practical application of Sen [52] and Nussbaum's [53] capabilities approach. In essence this identifies a framework through which the freedom to achieve well-being is linked back to human 'functionings'. These are divided into 'beings' and 'doings'; with 'beings' relating to the essence of the human, such as nutrition and education, and 'doings,' which are more active, such as voting and spending money. 'Beings' and 'doings' can be strongly linked but travelling as a function tends to fall into the category of 'doings' in terms of human 'functionings'. The capability approach focuses upon a person's actual ability to attain a functioning. This can be achieved by a 'conversion factor', the potential to turn a resource into something that supports a functioning.

Sen uses the example of a bike to develop the relationship between a resource and its conversion factor. [54] The ability of an individual to convert the resource of a bike into the functioning that is mobility depends upon a number of factors. These are environmental, the individual's physical surroundings that may or may not support bike riding; societal, the rules and regulations that may or may not permit this individual's ability to ride the bike; and personal, is the individual capable, both physically and mentally of riding the bike? A disabled person, due to, for example, societal barriers and individual impairments may not currently be able to drive a car and achieve the functioning that is independent travel. Autonomous car technology is part of a conversion factor; a resource that could, if technical, regulatory and legal systems permit, give disabled people the actual ability to attain this functioning.

6. But don't driverless cars have drivers? Are driverless cars driverless?

Having located the need for equality of access within the relevant ethical and societal framework there is a need to address the law to determine the extent to which it supports use of the technology in this way. Driverless cars are autonomous in the sense that they can drive themselves. Clearly, as they do drive around, without human intervention, something non-human is driving them. If they did not have drivers they would not go anywhere, let alone down motorways or around ring roads. So, it is argued, it cannot be accurate to call them 'driverless'. They have drivers, it is merely that those drivers are not human.

Given the modern usage of the term 'driver' as in 'printer driver' for example, it is consistent with colloquial speech to talk of non-human entities driving things, such as nails, gear wheels and printers. Indeed, the Concise OED definition of 'driver' is 'a person or thing that drives something'. As a definition for 'drive' it gives: 'Operate and control the direction and speed of a motor vehicle'…or 'Convey in a car.' [55] In either case, the sensors and programmes which operate and control an autonomous car are, technologically and linguistically, drivers. The full OED entry for driver has not yet been revised and is still largely as per the 1898 first edition. The first meaning for driver there is 'one who drives' and the third is 'A tool or appliance for driving.' The software 'driving' an autonomous car can be defined as 'a tool or appliance for driving' and therefore a driver. As stated above, autonomous cars do have drivers; but those drivers are not human beings.

This point is important because in order to use autonomous cars to address the injustice of access to personal mobility described earlier they need to be truly autonomous in the sense Horwick and Siedersberger [56] give:

'autonomous driving assistance systems (ADAS) …allow the driver to be distracted while the A-DAS is controlling the vehicle. This allows the driver to follow up side occupations like e.g. watching TV or writing emails. As such ADAS are able to realize self-determined action without any driver support, this is the only group that can be characterized as autonomous.'

Only with this level of autonomy can a partially sighted person, or a person with mental or physical impairments preventing them from safely controlling a non-autonomous car, rely on an autonomous vehicle to give them mobility independent of having another human in the car with them. A person unable to use a regular car could rely on such an autonomous vehicle independently. Horwick and Siedersberger base their analysis of the legalities of this technology on a key international instrument stating:

'According to the Vienna Convention in[sic] 1968 these DAS [Driving Assistance Systems] are not yet allowed by law, because it is postulated that 'every driver shall at all times be able to control his vehicle.''

This should not be a problem for an unmanned, in the sense of not carrying a human being, self-driving car if these autonomous cars are deemed to have drivers, and those drivers are 'at all times able to control' their vehicles. There is, therefore, a need to determine how the technical definitions pertaining to autonomous systems relate to existing legal provisions governing road use.

The Geneva Convention on Road Traffic [57] was opened for signature on the 19th of September 1949 and entered into force on the 26th of March 1952. Its aim was to encourage harmonised standards to govern the development of road transport systems and, in this way, to develop safe driving conditions. By 1957, both the UK and the USA had ratified this convention. This instrument was followed by the Vienna Convention on Road Traffic [58], which came into force on the 21st of May 1977. The UK signed this later convention in 1968 but has not ratified it, while the USA is not a signatory. In this way the international framework includes the Vienna Convention with over 70 ratifications and the earlier Geneva Convention which continues to bind States such as the UK and the USA.

There is, therefore, a need to examine in greater depth key definitions within the Conventions. The Geneva Convention defines 'driver' as:

'any person [our emphasis] who drives a vehicle, including cycles, or guides draught, pack or saddle animals or herds or flocks on a road, or who is in actual physical control of the same' [59]

Furthermore:

'Every vehicle or combination of vehicles proceeding as a unit shall have a driver.' [Drivers] 'shall at all times be able to control their vehicles or guide their animals. When approaching other road users, they shall take such precautions as may be required for the safety of the latter'. [60]

The later, Vienna Convention defines these terms in a similar but narrower manner:

'Article 1: Definitions, Section (v)

"Driver" means any person [our emphasis] who drives a motor vehicle or other vehicle (including a cycle), or who guides cattle, singly or in herds, or flocks, or draught, pack or saddle animals on a road'

Moreover, this person must fulfil certain criteria:

'3. Every driver shall possess the necessary physical and mental ability and be in a fit physical and mental condition to drive.

5. Every driver [our emphasis] shall at all times be able to control his vehicle or to guide his animals.'

These definitions, developed before the technology behind today's autonomous cars was developed are, as such, anachronistic. Furthermore, they have an unjust impact as in order to use an autonomous vehicle, which has a perfectly good non-human driver, a partially sighted or otherwise excluded person would still have to employ another human being to sit passively in his or her car in order to comply with the relevant legal framework. This is no better than the existing situation in which they must pay a taxi driver and therefore relies upon the agency of another.

Due to the outdated nature of these provisions in the light of technological development a March 2014 amendment to the 1968 Convention was passed with a new paragraph 5bis inserted into Article 8 (5) stating at subsection (b):

Vehicle systems which influence the way vehicles are driven and are not in conformity with the aforementioned conditions of construction, fitting and utilization, shall be deemed to be in conformity with the first sentence of this paragraph and with paragraph 1 of Article 13, when such systems can be overridden or switched off by the driver [our emphasis]. [61]

The updating of the law to reflect technological progress is to be commended. However, while this amendment paves the way for implementation of autonomous cars it still mandates the need for a human driver who has the ability to control or shut down the technology. For the purposes of the full participation of disabled people, this does not bring about effective change as it persists in envisaging the intervention of a human agent and does not support a truly autonomous technology.

Smith [62] provides an analysis of the legal situation in the United States, a country which, as outlined above, is not a signatory to the 1968 Vienna Convention, only to the looser 1949 Geneva Convention on Road Traffic. Furthermore in the United States each State makes its own rules about driver licensing. Already three States, Nevada, California and Florida, refer specifically to autonomous cars, however, in each case they require a human driver to be present and able to control the vehicle at any time. A general consensus has emerged that in essence autonomous cars are 'probably legal,' in Smith's terminology provided that:

'a human is able to intervene in the automated vehicle's operation.' [63]

This has generally been operationalized in automated car designs as requiring that there is a human being seated in the usual driver's seat, with access to the usual controls who can reclaim total control from the vehicle at any point. Professor Paul Newman heads a team of researchers based at Oxford University who are developing autonomous systems. The team emphasizes that there will always be a human driver available, [64] indicating that they take this requirement very seriously. Similarly the creators of Leonie, a system developed at the University of Brunswick Institute of Technology, refer to the Vienna Convention articles 1 and 8 requirements and note:

'Consequently, the driver has to be a (legal) person and not a robot. He will also be legally responsible for the vehicle guidance and compliance with traffic regulations. …Therefore, 'Leonie' must still have a specially trained safety driver who can always gain control of the car and monitors [sic] the traffic flow.' [65]

This position does nothing to address the injustice in access to mobility of present vehicle and infrastructure designs. Potential solutions to address this issue include: (a) attributing legal personhood to the (software) drivers of autonomous vehicles or (b) redefining 'driver' in the Vienna Convention and Geneva Conventions to include, for example, 'a person or thing which drives a motor vehicle.' Horwick et al appear to lean towards the former:

'In other words the vehicle has to act autonomous [sic] in terms of Immanuel Kant, who defined, that autonomy means self-determination according to the moral law. In the civil area the effective moral law are road traffic regulations, which aim for collision avoidance and physical inviolability.' [66]

Attributing legal personhood to the software systems driving autonomous cars would avoid changing the conventions but it would also raise questions about the distribution of responsibilities to entities which, fundamentally, do not respond to blame and praise in the same way as humans do. As these concepts lie at the root of our legal systems, especially tort and liability laws, this option might prove difficult to implement. The situation is further complicated by the numerous potential layers of human input into the technology's development with developers, programmers, mechanics and the passenger him or herself potentially incurring liability. This, to an extent, could be addressed with the creation of industry-specific international standards in a manner mirroring developments in healthcare robotics. [67]

Taking an alternative perspective, work in machine ethics [68] suggests that modern software systems can be programmed to follow rules of the sort found in human legal and ethical systems, with some caveats. However, such systems cannot be programmed to know what to do in areas where the human rules are indecisive or undefined. In particular, in those areas where our best ethical theories still disagree. Paradigm examples are so-called 'Trolley problems' in ethics. As the name might suggest, driving, especially in Europe, is a situation in which analogies of trolley problems can be expected to occur occasionally. With humans we allow freedom for conscience in these cases, but as it is not yet clear what a machine substitute for conscience would be this poses a problem for autonomous vehicles. One solution for this is that the vehicles are designed so as to be consistent with one or the other ethical theory, but the responsibility for this ethical choice remains with the human purchaser who chooses these ethical characteristics of their machine. [69] In this case it has been suggested that the machines would be 'ethical proxies' for their human owners, rather than moral agents or moral patients. If these issues, which lead us towards the wild frontiers of attributing moral status to non-human entities, seem too dangerous, perhaps the debate should return to the second horn of the dilemma.

The second horn of the dilemma involves changing the Vienna Convention's definition of driver to include 'persons or things' which drive. As outlined above, The Vienna Convention has been amended to address technological development but this does not support a fully autonomous technology. There is, therefore, a need to revisit the debate around this legislative change to clarify the relationship between the human and the autonomous system. The Geneva Convention has succeeded in retaining its original form so far. Given the vast change in transport technologies since its conception, and its even older origins in the treaties of the 1920s, perhaps it is not so radical to consider a wide-ranging amendment now, given the strong ethical justification for making the change.

There is a potential third option, which may be combined with either of the other two, though it raises ethical issues of privacy. We might wish, initially at least, to facilitate the use of autonomous vehicles by providing a means of oversight in which the equivalent of air traffic controllers are able to intervene and bring a vehicle to safety should its autonomous systems fail and its human occupants be unable to rectify the problem. Horwick and Siedersberger [70] propose technologies for safely managing autonomous cars which would bring them to a standstill on the roadside in a safe place. Non-autonomous cars already have this capability in some circumstances such as their use by taxi-cab firms. [71] It needs to be determined whether if autonomous vehicles employ similar technology but perhaps slightly more often, this is acceptable to the populations who might benefits from this technology.

Questions need to be raised around whether changing the concept of driver is really a high price for the rest of us to pay in order to enable all people to share equally the advantages of access to personal mobility and whether this is against the spirit of the Convention. Indeed, the Vienna Convention also requires:

'Drivers shall show extra care in relation to the most vulnerable road-users, such as pedestrians and cyclists and in particular children, elderly persons and the disabled.' [72]

This indicates an awareness of the need to balance the interests of diverse road users which could be drawn upon to support arguments in favour of the revision of key definitions.

7. Conclusions

In this paper certain assumptions behind the definition of 'driver' in the Vienna Convention have been challenged in the light of changing technologies of vehicle control. This has been done in order to address the injustices in access to personal mobility which the automobile introduced and to avoid similar injustices being built into the development of autonomous driving technologies.

Inclusion and access must be at the heart of the development of autonomous driving technologies in order to avoid further unacceptable and unethical discrimination against those with unusual body types or disadvantages from age or injury. These groups are particularly reliant on motorised transport for the mobility and independence they require to contribute fully to society. The Vienna Convention notes that drivers shall allow extra care in relation to the most vulnerable road users. It is argued that this also applies to the designers and regulators who control technological developments on our roads. They do have an obligation to those vulnerable to discrimination and lack of access to independent mobility.

The present definition of 'driver' as a person who drives precludes efficient access to these technologies by the groups with the greatest claims to benefit from them. Two options for addressing the resulting injustices have been outlined. One would be to designate the software systems which control autonomous vehicles as persons. A second would be to change the wording of the Vienna convention to 'persons or things'. The latter is recommended.

The analysis presented indicates that there are sufficient ethical reasons to justify changing the wording of the Vienna Convention to allow non-human road vehicle drivers; it is recommended that this course of action is followed. Questions have also been raised around how liability and ethical responsibility might be managed in ethically difficult driving situations if road vehicle drivers were non-persons. The application of the concept of 'ethical proxies', in which software can be programmed to behave in ways consistent with a given ethical theory has been evaluated. Vehicle owners or users may retain the ability to choose the ethical theory most consistent with their own consciences. The advantage of this is that it keeps the liability closest to where it lies today, with the human who is using the vehicle to achieve their goals. There is a need to base any legislative responses within such a broad ethical framework to enable consistency and clarity to be embedded within any reforms at the early stages of their development.