Self-driving but guided by people

Final report published by Reed Mobility

In March 2022, ITS UK member, Reed Mobility, was awarded funding by the Rees Jeffreys Road Fund as a winner of their competition commemorating 150 years since the birth of their founder, William Rees Jeffreys. Entrants were asked to describe what they would do to ensure roads could be enjoyed by all fifty years into the future, in line with the vision of Rees Jeffreys.

Envisioning a significant role for self-driving vehicles on the roads in 2072, Reed Mobility’s proposal sought to determine how society’s interests could be captured and integrated into the development and deployment of this technology. This builds on recommendations made to the European Commission by an expert panel (of which Reed Mobiilty’s founder, Nick Reed, was one) on the ethics of connected and automated vehicles (Bonnefon et al., 2020).

The project team, led by Reed Mobility, comprised DG Cities, TRL, April 6 and Humanising Autonomy and were supported by an advisory panel of international experts from the public, private and academic sectors. Focusing on a use case of self-driving buses operating in an urban environment, a survey asked participants to rate their agreement with numerous statements related to the behaviour and operation of such vehicles, for example “I would be happy for a self-driving bus to take more risks to catch up time if had been delayed by a traffic jam”. Each statement was associated (positively or negatively) with a range of societal values, such as trust, safety, legality and urban design. By analysing the data from more than 2,000 respondents, it was possible to estimate which of these values participants felt were most important
The survey was followed up by two workshops with members of the public held at TRL’s Smart Mobility Living Lab. Led by DG Cities, these delved deeper into the key outcomes from the survey, seeking to identify ‘ethical red lines’: the essential features of self-driving buses adoption that must be observed for their deployment to be considered acceptable. These included that such vehicles should not make safety worse, should operate within a clear legal framework, should not take risks to save time or to reduce cost and should seek to protect all road users equally.

The report suggests that societal values can be captured and help to feed into the design, development and regulation of self-driving vehicles. The authors recommend that using virtual reality tools to present self-driving vehicle behaviours may be an even more effective technique for exploring this issue. With transport technologies based on artificial intelligence emerging, the risk of misalignment between such advanced systems and the expectations and preferences of society is growing. The authors conclude that Rees Jeffreys’ vision for our roads to deliver prosperity and enjoyment for future generations will only be fulfilled by integrating societal interests into the way mobility is achieved. That sounds like intelligent transport systems to us!

The full report is available for download here     

About the Project

In 2020, the European Commission published recommendations on automated vehicle (AV) ethics . In follow-up, Reed et al. (2021) noted that the artificial intelligence (AI) systems controlling an AV could not derive the ethical principles that sit behind the behaviours they display, suggesting that AVs should be guided by overarching ethical goal functions (EGFs). These create a framework, governed by regulators, to embody societal expectations of AV behaviour and should be determined by an inclusive, deliberative process with the communities in which AVs operate.

The Rees Jeffreys Road Fund commemorated 150 years since the birth of its namesake with a competition to explore what roads will be like 50 years from now. For this competition, Reed Mobility submitted “Developing ethical goal functions for road use”, proposing to develop the process for engaging with communities to establish societal preferences that underlie EGFs and work with an AV technology developer to demonstrate how EGFs guide AI systems.

In March 2022, Reed Mobility was one of two winners of the competition and was awarded a grant of £75,000 to deliver this project.