Of the potential benefits of driverless vehicles, the one most consistently anticipated by witnesses and submitters to this inquiry is that of improved safety features and therefore improved safety outcomes. The Committee heard that this stands as the most likely and most important benefit of automated and driverless vehicles. Noting that over 1 200 Australians die and over 37 000 are injured as a result of accidents on Australian roads each year, the Committee considers these safety benefits to be of paramount importance in any discussion of the social impacts of driverless vehicles.
This chapter outlines the major benefits to road safety likely to emerge as a result of increasing automation of vehicles, as well as the further consequences of those. It also canvasses some of the concerns raised in this context, including the ethical questions arising from accidents and other incidents involving automated vehicles, as well as the issues raised during period in which there is a mixed fleet on public roads.
The Committee heard from a wide range of submitters and witnesses to this inquiry that the safety benefits of increasingly automated vehicles will be the technology’s most significant outcome.
As the Department of Infrastructure and Regional Development (DIRD) submission noted:
Road crashes in Australia kill about 1,300 people per year. The cost of road crashes to society has been estimated to be $27 billion annually, or 1.8 per cent of GDP (based on a willingness-to-pay methodology of valuing human life) (Transport and Infrastructure Council, 2016a) (BITRE, 2014b, p.28). International evidence indicates that human error may be a factor in more than 90 per cent of crashes, and that road user distraction or inattention is a contributory factor in around 10-30 per cent of road accidents (Singh, 2015), (TRL, TNO and Rapp Trans, 2015, pp. 54-55). This does not necessarily mean the driver is the cause of the crash; however, it does indicate that human error may be the predominant factor in most road accidents.
If automated technology reduces or eliminates human errors, as is generally expected, then benefits for road safety may be substantial.
Mr Christensen of the iMOVE Cooperative Research Centre, when asked about the potential benefits of autonomous vehicles, argued that:
The largest benefit altogether will be in safety and the avoidance of accidents and road trauma and the costs that are associated with that. It is a massive burden on the economy and the prospect for autonomous vehicles to substantially reduce the rate of accidents and trauma, therefore, stands to benefit the economy in a dramatic way.
The Australasian College of Road Safety (ACRS) argued that, not only are the direct effects of road trauma serious, they represent only one element of the consequences:
With 25 people dying and 700 being seriously injured each week in Australia, the ripple effect of each road trauma event to our families, to the workplace and communities is enormous. It is reasonable to assume the cost to the national economy over the next decade to be in the order of at least $350bn.
The subsequent impact on Australia’s health system and communities is too often overlooked, as is the impact on national productivity.
The Road Trauma Australia 2016 Statistical Summary, published by DIRD in July 2017, notes that the annual reduction in Australia’s death toll over the past decade has been reversed in the last three years, with annual increases across all states and territories.
A significant proportion of the serious road trauma on Australian roads occurs in regional and remote areas. The National Farmers’ Federation argued that autonomous vehicles should improve safety ‘by removing the risks of driver error and driver fatigue during long and tedious travel on country roads’.
Mr Lauchlan McIntosh of ACRS pointed to Australia’s fall in road safety results, noting that a focus on improving Australia’s performance has not been enough of a priority:
In our road safety performance we have fallen from about ninth to about 19th. We used to be at the top of the game. We were in the top 10. We are now in the best 20. That is not good enough. Why is that so? We have really let ourselves slip. We need to do this not just by ourselves. We need to do this with other countries. We need to sign into the US and we need to sign in to Europe and be part of that. ANCAP is doing that with our merger with the Europe ANCAP testing, but it needs to happen in all these other areas.
Austroads also noted the significant levels of road trauma in Australia and the potential for autonomous vehicles to reduce those levels, while also highlighting that even vehicles that do not meet the definitions of autonomous should contribute to this:
Perhaps the most significant societal benefit that could be achieved with technologies that automate the driving task is the potential to significantly reduce road trauma. As you have already heard today, road crashes lead to about 1,300 fatalities per annum and over 30,000 serious injuries each year. It has already been highlighted by others that studies have estimated that over 90 per cent of road fatalities have human error as a causative factor. The potential for automation to reduce this figure is significant; however, it is not just driverless vehicles that can address this issue. Vehicles with self-driving capabilities that are not necessarily driverless also have significant potential to mitigate human error with the driving task and thus improve road safety significantly. Austroads would like, therefore, to suggest that any recommendations from this inquiry relating to road safety be not limited to just vehicles that are driverless.
Pointing to existing technology, manufacturer Volvo noted that the incorporation of SAE level 1 or 2 into some vehicles has already demonstrated improved safety performance:
Research conducted in the US highlights the value of cars equipped with safety features that would be standard in fully autonomous cars. The Insurance Institute for Highway Safety’s (IIHS) 2016 survey found that that cars equipped with front crash prevention technology are much less likely to rear-end other vehicles.
In the first study of the feature's effectiveness using U.S. police-reported crash data, IIHS also found that cars with automatic braking reduce rear-end crashes by about 40 percent on average, while forward collision warning alone cuts them by 23 percent. The autobrake systems also greatly reduce injury crashes. The rate of rear-end crashes with injuries decreases by 42 percent with forward collision warning with autobrake.
IIHS concluded that If all vehicles had been equipped with autobrake that worked as well as the systems studied, there would have been at least 700,000 fewer police-reported rear-end crashes in 2013. That number represents 13 percent of police-reported crashes overall. Front crash prevention would be a standard safety feature incorporated into fully autonomous cars.
Similarly, representatives of the New South Wales Government told the Committee that these existing technologies are having an impact on road safety:
So already in the market, with registered vehicles operating on our roads in New South Wales, we have the benefits of those technologies which, when we look at the research that is being done, are already contributing significant reductions. When we see that rear-end crashes and when we see head-on crashes—crashes into objects on the road constitute more than 60 per cent of the fatalities that we have, and a significant amount of the 12,000 serious injuries where people are admitted to hospital in New South Wales each year. We can see that those technologies are already having a benefit when comparing those vehicles against vehicles without the technology. So right now we already have the benefits coming through, which are all clearly technologies that have been developed against the vision of having driverless vehicles in the future.
As Mr James Goodwin of ANCAP argued, the increasing automation of vehicles will have a significant and important effect in improving road safety: ‘with more than 90 per cent of crashes involving human error, automation is really the key in reducing road trauma, and the future of vehicle safety lies with these active and autonomous safety features’.
As such, from 2018, ANCAP’s assessment of new vehicles will take into account automated driving technologies: as of that year, only vehicles with some level of autonomous technology included as standard can be rated as a five-star ANCAP vehicle.
While the safety of vehicle passengers has been the main focus of this chapter, the Committee also heard that increasingly automated vehicles will have implications for the safety of non-passengers. However, while there was general agreement that autonomous vehicles would lead to improved safety for passengers, some witnesses expressed concern that non-passenger safety has not been adequately considered in the debate on driverless vehicles to date. In particular, the Committee heard concerns about the safety and rights of vulnerable road users – generally defined as pedestrians and cyclists – and the extent to which those may be affected by the increasing levels of automation in vehicles.
As Ms Katie Minogue of Maurice Blackburn Lawyers argued:
… it is really important that consideration is given to vulnerable road users in the regulation and governing of those programming decisions. It is crucial that manufactures of that technology are not automatically assigned the decision-making power. And it is really important that it is a conversation in which society also participates in and that vulnerable road users have a voice in also.
The Amy Gillett Foundation, advocates for road safety for cyclists, noted research suggesting that all road users will need to familiar with autonomous vehicle technology, to understand the capabilities of the vehicles:
We also draw your attention to recent work from the respected Dutch safety research institute, SWOV. This highlights the need to anticipate behavioural adaptations by vulnerable road users to an increase in automated vehicles. Pedestrians and cyclists are likely to appreciate messages from the vehicles that they have been detected and what action the vehicle is going to take. The form of these messages needs to be determined.
The submission from the University of the Sunshine Coast highlighted that, since human drivers are often unable to adequately respond to vulnerable road users, work remains to be done on how a vehicle should be programmed to do so:
In a recent study of road user behaviour at intersections, Salmon et al (2014) found that human drivers are not prepared for the variability in behaviour displayed by cyclists, motorcyclists and pedestrians. How AV designers can develop systems that a. understand this level of variability, and b. are able to cope with it in complex road scenarios, remains unclear.
Ms Cecilia Warren of the Insurance Australia Group (IAG) noted that recent evidence has shown that, while there has been an increase in the safety features available in vehicles, vulnerable road users remain part of the equation:
We have to remember as well that roads are not just used by vehicles but also used by pedestrians, bicyclists, vulnerable road users and others. Even in a fully autonomous scenario, it is still not just autonomous vehicles that are using these roads. As we head towards full autonomy, we are still looking at interactions of the human with the machine, to put it in those terms.
The NTC’s submission stated that the safety assurance system being developed will ‘look to assess the safety of the vehicle for both passengers – in a shared mobility or private ownership setting – and vulnerable road users’.
DIRD highlighted that all road users – passengers and others – should expect to see improvements to safety as a result of highly automated vehicles on the road:
If automated technology reduces or eliminates human errors, as is generally expected, then benefits for road safety may be substantial. The expected safety benefits of automated vehicles extend to other vulnerable road users, such as pedestrians and cyclists, since vehicles with higher levels of automation (i.e. SAE Levels 4 and 5) will be able to detect their presence and take evasive action automatically (Somers and Weeratunga, 2015).
However, DIRD also noted that expected safety benefits are at this stage only expected – in the absence of significant trials on public roads, actual consequences have not yet been seen:
Importantly, the expectation of near zero fatalities with highly automated vehicles may not be realistic, including for the duration of a mixed fleet... More trials and real-world experiences are required to understand the safety impacts of higher levels of automated driving.
While proponents of driverless vehicles argue that increased road safety is one of the primary benefits of increasingly automated vehicles, multiple studies have demonstrated that uncertainty around the safety of these vehicles is a key concern for many people. This ambiguity is increased by media reports of accidents involving automated vehicles, which highlight the new and unknown nature of the technologies.
Research by the Australian Driverless Vehicle Initiative found that 80 per cent of people surveyed expressed concern about ‘the ability of the car to perform safely in all conditions’ as one of their key concerns regarding autonomous vehicles.
Mr James Goodwin of ANCAP noted that new technologies, even those which improve safety, have often taken some time to gain widespread public acceptance, and for that reason it is important for vehicle manufacturers and other stakeholders to engage with community concerns:
People did not like air bags in the old days, and they did not like seat belts even before that. We have to learn from those experiences and know that this is the newest form of technology that we need to bring the consumers along for the ride. We have to make sure they have the confidence in it. We feel that the level 1 features we are currently evaluating and assessing are the building blocks to a full driverless future. We need to get that right now and we need to make sure that the systems work and that consumers have the confidence to use them and know how they do work to make sure that they know what is in their car and how it works.
The Australian context
The Committee heard that, while safety principles are largely universal, Australia faces unique challenges in ensuring passenger safety in autonomous vehicles. In particular, kangaroos present uncertainties unlike many other animals that may be expected near roads. Witnesses noted that this problem, which has received some media attention, will need to be solved before passengers will feel safe in autonomous vehicles on Australian roads.
Mr David Pickett of Volvo Australia discussed the difficulties which kangaroos present for autonomous vehicle software:
We have done a research project here in Australia for kangaroos, which has proven to be very interesting. The cars at present do not take what you would call a drastic measure. They will brake. At the same time, though, we have technologies in the car so that, if we do brake hard for something, the emergency brake warning lights come on on the car. It will start flashing the hazard lights, and the high-mounted brake light will flash to warn the other drivers earlier. The decision about what we do for kangaroos has not been made. At this point in time, everyone is concentrating on reducing the speed. If we can knock off 20 kilometres an hour or 10 kilometres an hour, it might be the difference between driving home with a busted headlight or walking.
The regulatory approach
The Department of Infrastructure and Regional Development outlined the approach to safety being taken by the COAG Transport and Infrastructure Council to develop a national performance-based safety assurance regime:
In November 2016, the COAG Transport and Infrastructure Council agreed that a national performance-based assurance regime should be developed to ensure the safe operation of automated vehicles, in line with international practices (NTC, 2016b). Such a safety assurance regime will give consideration to issues such as an automated vehicle’s ability to obey speed zones and traffic controls, interact safely with other road users and perform safely in all likely road and environmental conditions (NTC, 2016b).
Mr Marcus Burke of the NTC expanded on this approach, noting that it places the burden of demonstrating safety assurance on the vehicle manufacturers:
I think this goes to the approach we are looking to develop in terms of both guidelines and broader safety assurance. We are trying to learn from rail and aviation and moving away from the prescriptive approach that we in the road sector have had towards more of a safety management approach. What we are envisioning that would look like—again, initially for trials and then potentially for broader deployment—is the operator of these vehicles coming to government and demonstrating that they have identified all the key risks based on the scope of operations that they are looking to operate in and they have systems and processes in place to mitigate, eliminate or manage those risks. That will be dependent on the scope of operations. […] So we are looking to try and take an approach which does not prescribe particular technologies or a mix of technologies, or assume a particular business model; it is about putting it back on industry to demonstrate how they have managed the risks. We believe that is the right approach both for trials and then for broader regulation to ensure that we are not putting in regulation that is going to become outdated very quickly and that does not assume a particular technology that will finish up not being used.
Driverless vehicles will pose ethical questions in ways that vehicles with human drivers do not. The fundamental question is how autonomous vehicle algorithms will be designed to protect the safety of passengers and non-passengers. As the Committee heard, resolution of these questions is central in encouraging public acceptance of driverless vehicles on public roads.
The well-known philosophy exercise ‘The Trolley Problem’ lies at the heart of this question. The Trolley Problem, in its basic form, revolves around a scenario wherein a vehicle can either stay on course and injure a number of people or divert, harming a smaller number of people, inviting people to consider under which conditions they would divert the trolley. As applied to driverless vehicles, the question becomes: in a scenario where people will be harmed regardless of what the vehicle does, what action should the algorithm controlling the vehicle take?
This situation, as the submission from the iMOVE CRC noted, ‘is not a new dilemma, but it is one that is being highlighted frequently in the media and which requires some attention in order for the general public to feel sufficiently safe that they are willing ‘grant’ the technology a ‘social licence’ to operate’.
The submission from the Queensland University of Technology similarly noted that this issue, and how it is handled and resolved, will affect how people think about autonomous vehicles:
Part of the disquiet around automated vehicles is an emotional reaction to the inability to understand their decision-making or predict their behaviour. Ignorance of the perceptual and decision-making systems of automated vehicles risks them being judged as negligent or reckless agents or even a public menace. For example, what issues arise for acceptance in terms of regulation and consequences for socially unacceptable behaviour (e.g. driving too quickly in a car park and frightening humans, even if the vehicle is rated as safe moving at those speeds)? We are still at an early stage of developing our understandings of public perceptions of and interactions with automated vehicles. Consumer engagement in the development of programming decisions about vehicle responses in a crash or emergency will be an important aspect of ensuring consumer confidence in this new technology.
The ARC Robotic Futures Research Team also made the point that these ethically difficult decisions will need to be made on the basis of transparency and public debate: ‘It cannot be a hidden process; it needs to be out in the open and publicly debated’.
As DIRD noted, the United States Department of Transportation draft Automated Vehicle Policy from 2016, ‘suggests that manufacturers address ethical issues consciously and transparently with input from various stakeholders’.
One issue raised by witnesses and submitters was concern that insufficient attention has been paid to date on the reality that, even once highly automated vehicles are commercially available and legal on public roads, such vehicles will be in a minority. Australian roads will feature, for many years to come, a mixed fleet, consisting of vehicles of markedly different levels of automation and safety features. As multiple experts emphasised, this will have repercussions on how road safety is managed.
Mr James Goodwin of ANCAP noted the extent of this problem:
We have an average vehicle age at the moment of 10 years, and older cars are overrepresented in fatality crashes. With one in five cars on the road more than 15 years old, it is going to be a long time until we get this technology rolled out across the fleet. What we are saying is that we really need to protect everyone on the road. Hopefully cars will start avoiding crashes in the first place, but there will still be some crashes in the future. We are going to have to understand how all those different vehicles will mix on the roads.
Mr Ashley Wells of the Federal Chamber of Automotive Industries also pointed to the average age of passenger vehicles in the Australian fleet, and noted that this fact in conjunction with the density of vehicle ownership in Australia means that, even once highly automated passenger vehicles become commercially available, there will be a long period where driverless vehicles share the roads with fully driver-operated ones.
Access to data in the event of an accident
As noted in chapter 2 of this report, the use of data created by driverless vehicles is a key point requiring clarification to encourage public acceptance. The question of data ownership and access becomes of particular importance when an autonomous vehicle is involved in an accident which causes harm to passengers, other people or to property.
This data would be able to serve multiple purposes. It would be able to provide an accurate and precise reconstruction of the circumstances of, and reasons for, the accident, thus clarifying liability issues. It would also provide manufacturers with the capacity to improve the model’s safety performance by assessing any shortcomings which the accident revealed.
Ms Cecilia Warren of the IAG summarised the main reasons why this question is important:
At the moment, as insurers, when a collision occurs there is a lot of reliance on human testimony. Experts are used to understand or potentially reconstruct a collision. What the promise of future vehicles provides is that there will be data that will help to describe what happened, why it happened, where it happened and who was involved. There are a couple of things that are really important with that. Firstly, ensuring that similar incidents do not occur again is a really key piece around ensuring that we learn. We know from other industries, such as aviation et cetera, that there will be things that go wrong that we cannot predict. The key to that is a continual learning piece. Secondly, there will be a need to determine, when something going wrong, who needs to recompense and where liability needs to fall. Finally, there are concerns around both ethics and privacy for the consumer. Is it the consumer's data? Is it the manufacturer's data? Is it the government's data? All those questions are currently being grappled with by the National Transport Commission, and beyond that as well.
Ms Katie Minogue of Maurice Blackburn Lawyers argued that public acceptance of driverless technology will require that people understand what data is collected by vehicles and what access they will have to that data in the event of an accident:
… consumers need protection and guarantees in terms of how their data is handled and kept, that there be standards on the use of that data and that in the event of an accident they not be denied access to relevant event or crash data that is going to help determine liability in an accident. Certainly consumers are going to need certainty and appropriate support and care if they are involved in an accident that involves an automated vehicle. It is really important that people have confidence in that in terms of their acceptance and uptake of the technology.
The National Transport Commission (NTC)’s work program for autonomous vehicles indicates that in 2018 it will undertake a project on ‘regulatory access to C-ITS and automated vehicle data’, with an aim to:
Develop options to manage government access to automated vehicle data that balances road safety and network efficiency outcomes and efficient enforcement of traffic laws with sufficient privacy protections for automated vehicle users.
As noted above, many witnesses highlighted the positive impacts expected to emerge as a result of the improved safety features of highly automated and autonomous vehicles. However, as with other aspects of driverless vehicles, there will likely be further and broader consequences of this improved level of road safety.
For instance, witnesses noted that considerable policing resources are currently devoted to road safety issues. As vehicles become increasing automated, there will be a corresponding decrease in the need for this work. In the longer term, this would affect the employment patterns of the police force.
Further, the current road trauma levels are a substantial component in the work of health care professionals, particularly in emergency departments of hospitals. If, as many have argued, increasingly automated – and eventually driverless – vehicles lead to a significant decrease in the number, and severity, of road accidents in Australia, then the work of those in the healthcare sector would correspondingly shift.
Another sector likely to change as a consequence of increasing automation of vehicles is vehicle repair – mechanics, panelbeaters and similar occupations – where a decrease in the number of road accidents would lead to a reduction in the need for people to carry out those repairs.
Given the wide range of witnesses arguing that autonomous, or even highly automated, vehicles could lead to a substantial reduction in the number of deaths and injuries on Australian roads, the Committee is of the view that this important social goal should make the introduction of these vehicles a priority for Australia.
The Committee notes that public engagement remains a key component of the move to autonomous vehicles, and emphasises the importance of stakeholders, including Commonwealth and state and territory governments along with vehicle and software manufacturers, conducting open public dialogue regarding safety concerns with driverless vehicles.
The Committee is of the view that the approach to driverless vehicle ethical questions suggested in the United States’ policy should apply broadly, and that manufacturers should be transparent in how their vehicles will behave in ethically difficult situations. Such an approach will help to ensure that Australians are comfortable with the automation of these ethical difficulties.
The Committee considers that an adequate regulatory framework for the ownership of and access to vehicle data is one of the key issues affecting public acceptance of driverless vehicles. As the next chapter will discuss, uncertainty around legal liability in the case of an autonomous vehicle accident remains one of the major concerns people hold about the technology, and clarity around the data generated during such an event is an important component of resolving the issue.
While these issues remain distant contingencies at the present time, and dependent on the changes to road safety expected by many experts, the Committee believes it is important that these consequential impacts of increasingly automated vehicles are discussed and understood as part of the broader discussion of the topic. The introduction of driverless vehicles, when it happens, will have substantial and yet to be understood impacts on many aspects of life. These changes should be discussed as part of broad public engagement on the topic of driverless vehicles, as recommended in chapter eight of this report.
The Committee recommends that the Commonwealth Government further investigates the issue of data rights for consumers, vehicle manufacturers and third parties such as insurers and relevant government agencies.