Choices

In creating a new legal framework capable of accommodating cases in which there is no driver to prosecute, the aerospace industry may offer a useful model to follow.

In instances of air disasters in which pilot error is ruled out, responsibility lands on those responsible for the design, manufacture, and maintenance of the systems in question.
This is not an entirely novel approach for ground-based vehicles. The previously mentioned ‘Vision Zero’ model, which has been widely adopted in many countries, considers road designers and vehicle manufacturers to be as important to road safety as the person behind the wheel.

This may have moral shortcomings. The designers of AI based on deep learning principles cannot fully understand or predict the outcome of these systems (Miles, 2021). Risk assessment becomes progressively more difficult as the systems increase in complexity and reliability.
To put it in plainer terms, the safer we make the self-driving car, the more difficult it is to know with certainty just how safe it is.
As the increasing complexity of deep learning models is likely what will make the self-driving systems of tomorrow safer, it would be unreasonable to prosecute creators for failing to address fatal flaws that might take billions of hours on the road to uncover.

If serious road vehicle crashes become as infrequent as air disasters, it may even be necessary for special road crash investigation units to be set up, analogous to those that investigate air disasters.

Whilst it sounds like Asimovian science fiction, we cannot completely rule out the possibility that future AI drivers may be intelligent agents with the legal responsibilities of a person. What to do in this circumstance should remain a question for future ethicists with better understandings of these new ‘persons’, but a retreat to the familiar legal territory of the 20th century is plausible. Unlike truly autonomous systems, Level 3 automation represents perhaps the toughest challenge for lawmakers. Treating them as if they were a Level 4 or 5 system invites abuse by users of ‘Level 3s’, who may try to pin their own mistakes on the machine.
Without appropriate legislation, consumers may feel uncomfortable using their Level 3 system if a mistake by the system becomes their problem.
In both cases, manufacturers are incentivised to ensure that their ‘Level 3s’ can provide impartial and indelible logs.

Transition and Integration

The start of the self-driving vehicle era will not mark the end of the human driver. Even places that seek to transition to a purely self-driven fleet as quickly as possible will have a transition phase.
How this transition phase is approached represents one of the most significant choices that governments will have to make. Rolling out city-by-city is the approach taken by Waymo and Baidu, and has several clear advantages.

  • Discussions with local governing bodies can be fronted by the company’s most senior or capable staff, increasing confidence that concerns are taken seriously.
  • Marketing and awareness campaigns can be highly targeted.
  • Any quirks specific to the area can be ironed out and integrated into the AVs before expanding operations further.

Choices for New Zealand

New Zealand, like anywhere else, needs to be ready for the self-driving car if it wants to benefit from this emerging technology, and now more than ever represents a fork in the road.

One path forward, and the path of least resistance, is to merely observe overseas developments. Superficially this would cost the government little and allow New Zealand to capitalise on the benefits whilst shouldering far fewer of the risks.
Lawmakers would be able to follow and implement trends that work, whilst avoiding pitfalls of pioneers.
Consumers and other road users would have confidence that when self-driving cars do reach our shores, they are less likely to suffer from the unfortunate and perhaps fatal consequences of being an early adopter. Designing around increasingly available Level 3 self-driving cars would be the most pressing article, as they represent the most imminent change to New Zealand’s roads.

Whilst this represents the cheapest option on paper, being too slow to adopt self-driving cars might run a hidden cost. If AVs dramatically improve road safety, then everyday New Zealand is behind the curve, which could represent a genuine cost to government resources and human lives.

A more proactive path could see the government steer New Zealand towards becoming an early adopter of self-driving vehicles to address specific goals. This could be a way to shore up existing fragilities in the transport sector, such as the current shortage of bus drivers, or to reduce serious injury and death in road accidents to meet the 2030 target of a 40% reduction from 2018 levels.
This approach would require investment and active partnership-seeking with autonomous driving system manufacturers. These partnerships would take time to develop and may not bear fruit. However, such an approach would allow New Zealand to capitalise on any benefits that manifest at the first opportunity.

The innovative path would also be the most fragile, as it would be a long-term initiative requiring constant reinvestment. A government looking to tighten its purse strings could kill the project with little to show for it.
Putting the project on ice would not be an option. Self-driving technology is developing fast, and renewing efforts after a period of slumber may be completely unnecessary if tried and tested systems could be imported instead.
Furthermore, stakeholders would need to be tempted in from overseas, and they would be quick to leave for areas with more developed tech sectors if the project ceased to be sustained.