UNECE tackles philosophy’s trolley problem
Self-driving (autonomous) vehicles present us squarely with one of the toughest brain-teasing conundrums in philosophy and ethics: the trolley problem – do you allow your automobile to kill one person if it can save several more? Peter Hulm, who took a doctorate in modern philosophy, examines the answers from a working group of the United Nations Economic Commission for Europe (UNECE).
First introduced in its bluntest form in 1967, the trolley puzzler has been “used extensively in empirical research on moral psychology” since 2001, according to wikipedia. It’s also known as the fat man dilemma, because the question is often posed as: would you push a fat man next to you over the bridge and onto the road below to block a runaway trolley if it will save five people in its track?
Now you can turn to UNECE for practical guidance on the main “fat man” issues thrown up by self-driving buses of the sort running around the pedestrian streets of Sion, and increasingly in other parts of the world (including a World Health Organization meeting in Geneva in 2019). In contrast to the fat man conundrum, the questions UNECE considers are of vital importance in determining how countries apply rules for self-driving autos, and the challenges have already appeared in a number of accidents that have occurred involving autonomous automobiles of all kinds.
Who's responsible for safety?
On 3 September 2019 UNECE’s World Forum for the Harmonization of Vehicle Regulations approved a new Framework document on automated/autonomous vehicles (PDF). One big question is: should manufacturers and software engineers be responsible for making their vehicles safe?
The forum’s conclusion: “The level of safety to be ensured by automated/autonomous vehicles implies that ‘an automated/autonomous vehicle shall not cause any non-tolerable risk’, meaning that automated/autonomous vehicle systems, under their automated mode ([ODD/OD]), shall not cause any traffic accidents resulting in injury or death that are reasonably foreseeable and preventable.”
The six-page document includes list of issues such as fail-safe responses, human intervention, obstacle detection, validation for system safety, cybersecurity and an event recorder. An annex sets out a full programme of work priorities for the specialists, with deadlines.
Maintenance and inspection not included, not education
The paper notes that the group does not consider as immediate priorities topics such as vehicle maintenance and inspection, consumer education and training, crash worthiness (differences between driving a standard and automated vehicle) and post-crash behaviour. The working group – which included China, Japan and the United States as well as European nations – updated an earlier document this year that took country views into account. The group is part of UNECE’s Inland Transport Committee.
But say when an autonomous vehicle can be used
The World Forum states that “vehicle manufacturers should document […] the specific conditions under which the automated vehicle is intended to drive in the automated mode [and] should include the following information at a minimum: roadway types; geographic area; speed range; environmental conditions (weather as well as day/night time); and other domain constraints.”
UNECE summary: Safety at core of new Framework to guide UN regulatory work on autonomous vehicles. (LINK)
Update: ITU devising 'driving test' for automated vehicles. 6 November 2019. (LINK)
Earlier Global-Geneva coverage of UNECE: Speaking up for the other Europe.