Mobility and Access Committee for Scotland (MACS) - Response to CfE on Automated Vehicles (AVs) Regulations

Accessibility in AVs for disabled people

There are several potential issues when it comes to accessibility in AVs for disabled people. For example, how could a fulltime wheelchair user access the vehicle without assistance if we assume driverless vehicles? How does a visually impaired individual recognise the vehicle at a distance considering their acuity may be low? How does a cognitively impaired individual have full understanding and agency over their journey when it cannot be communicated to them in a comprehensible manner?

These factors simply don’t seem to have been considered, with the focus of these systems being on nondisabled individuals, with an underlying assumption of high technological literacy.

A transport system is not accessible if a person cannot access it with as much independence as possible. AVs could easily become yet another inaccessible system if regulations do not consider the needs of disabled people such as physical accessibility.

What impact would AVs existing have on disabled people for differing modes of transport?

Transport presently suffers from an existing lack of accessibility – buses can be unreliable if you are a fulltime wheelchair user, the subways are simply inaccessible, and taxis have also become unreliable. Cycling isn’t accessible for many disabled individuals either, so essentially people are left with over surface trains which suffer from a lack of accessible stations and staffing issues. AVs solve none of these issues and might just present a context in which funding goes towards infrastructure that leaves disabled populations out.

This lack of access will simply worsen the levels of isolation, and lack of community engagement that disabled people are already dealing with. A potential lack of funding to fix the existing issues will also worsen the existing problems with accessibility. Introduction of AVs doesn’t solve anything and may just create yet another context in which disabled people are unable to access something that becomes a presumed norm.

EQIAs and Equalities

The language should really reflect equity not equality. Equality of opportunity does not lead to equality of outcome, equity of opportunity (a fair opportunity) does.

The presumption of individual policies as opposed to a more global guidance places the responsibility onto individual companies, and individual companies are disincentivised to provide anything close to true accessibility as it costs money to implement this. This also positions the state as not responsible for accessibility, which seems in stark contradiction to the Equality Act itself.

The regulatory framework should bring clarity to the fact that the government has existing responsibilities under the Equality Act and outline the foundations as to how they intend to meet these responsibilities as opposed to the creation of individual policies that will in no way cover the baseline expectations for accessibility as set out in existing legislation. Any future Impact Assessments (IAs) should consider previous work that has outlined accessibility responsibilities and use their data to complement our existing knowledge bases.

Without a set of baseline expectations for IAs and building it solely off responses will guarantee that some impacts will not be built into the assessment. It seems short-sighted to ignore decades of vehicle accessibility understanding when discussing the introduction of a new transport technology, and to instead propose we start from scratch creates circumstances in which potential users of the technology will not be represented.

For bare minimum accessibility, the above which implies the future minimum will be derived from responses, as opposed to derived from existing vehicle accessibility guidance. As existing guidance isn’t adequate, anything lower than this could easily become simply inaccessible and therefore unacceptable to disabled users of AVs.

Language and Expectations

Language is important for clarity. The popular understanding of “autonomous” which can mean anything from lane checking technology to full driverless cars simply isn’t clear enough.

As for future expectations – the very language used within this discussion centres on the idea of “autonomy” when no true autonomous vehicles exist at this present moment. Consequently, there needs to be stricter and clearer understandings of what is being discussed – is it assisted driving? Is it high automation? The reason for this being important is because the presence (or lack thereof) of a driver directly impacts the accessibility of the vehicles for many within the disabled population. For example, how is an independent visually impaired cane user meant to access the vehicle when visual recognition of the car isn’t within their capability? How is an otherwise independent powerchair using traveller meant to access a vehicle that they cannot load their chair into without assistance? How is a cognitively impaired individual meant to trust that a car without a driver will get them safely to their destination?

These examples illustrate only a small selection of potential problems – and whether we are discussing high automation (no driver) or partial automation (often present in cars with drivers) is a significant factor in these potentialities.

Training and Awareness for Drivers, Companies etc.

Who is offering this training? There is a question that now becomes about profit – for example PAVE UK (Partners for Automated Vehicle Education) is through the University of Warwick and specifically wants to “make the UK the leader in artificial intelligence (IA). PAVE UK also describe themselves as “The country’s first initiative to enable public trust and acceptance for automated vehicles through public education and awareness programmes”– they list amongst their founders the UK Government for Department for Business & Trade as well as the Department for Transport.

This reads much like an educational pathway, training pathway, awareness pathway is being constructed through PAVE UK who have already decided that the answer to the use of AVs in the UK is a positive one.

Due to this seemingly uncritical acceptance of AVs, concern would be that training and awareness for drivers etc. may exclude any potential issues with the placement of the technology into society.

Safeguarding Issues

As above, if the decision to use AVs has essentially already been made how will potential safeguarding issues impact that?

Reed Mobility’s Nick Reed is cited as a founder of PAVE UK. Their website elucidates how they have something called Vision Zero which was initially approved by the Swedish Parliament in 1997. Vision Zero exists to attempt to create a road system with no fatalities or serious accidents.

We note there is nothing in that plan, or on Reed Mobility’s website pertaining to accessibility though there is a study in the navigation bar that identifies a need to engage with the public around self-driving vehicles. There is also some discussion of ethics within AI, but no discussion about safeguarding beyond attempting to the creation of discretion within the IA systems.

It also discusses the creation of “Ethical goal functions” (EGFs) and through a fairly dense text they outline a system in which AVs have constructed goal behaviours that have been deemed ethical. These goals are arrived at through surveying the public for “red line” critical principles i.e. actions that the system should never take.

Participants within the study conducted by Reed Mobility noted that:

“A high level legal and governance framework was deemed important by participants … to ensure that users are able to trust service providers and vehicle manufacturers.”

Safeguarding seems a much more complex issue than has been discussed in the CfE and requires more understanding as opposed to an implicit acceptance that those who may derive profit from the adoption of AVs will always give an unbiased answer.

Question 121: In your view, are there any wider considerations regarding accessibility that should be taken into account in the deployment of AVs?

Yes.

We have significant concerns regarding any potential safeguarding issues and how they would impact already vulnerable populations who are a) unable to access present academic literature around the ethical considerations of AVs/IA systems and b) may have to interact with systems that do not yet seem to be fully understood either by users, or by experts in other fields who seem to be working with a level of uncritical acceptance.

All domains of access – physical, cognitive, visual broadly – seem impacted by the potential usage of AVs and these could easily become yet another barrier for disabled people (similar to how a wheelchair user can struggle to get an Uber, never mind more traditional forms of transport i.e. buses in Glasgow). If a person cannot access something independently, it is not accessible – and we can’t foresee a situation in which a person with even a moderate impairment could independently and safely access a vehicle with a system that is not well understood and is not discussed in clear terms. 

The CfE discusses the creation of an Automated Passenger Services Accessibility Advisory Panel (APSAP). It strikes us however, that if disabled passengers simply can’t access the vehicles – due to driverless vehicles depriving them of necessary assistance – then this panel won’t be particularly effective. Although it also notes evidence was sought on the safety impact, no evidence is presented about how disabled populations will interact with this technology.

In conclusion, we feel that there has been a concerted effort to position AVs as broadly accepted as a good idea. However, even the research conducted by a founder of PAVE UK speaks to foundational problems within the systems themselves surrounding ethics, and a public lack of support for the adoption of the technology within UK society.

An illustrative quote from the study:

“As the world continues to become more automated, using ethical AI will reinforce human equality in our cities and urban mobility systems… People will trust AI more as they begin to understand how it can enable a better quality of life for them – and how the risks are minimised.”

This appears, like a technology looking for broad public acceptance. Not a technology that there is demand for – uncritical acceptance (as present throughout the CfE) is not a position that is amenable to safety, nor accessibility. The most pressing wider consideration is the question of is this technology necessary? If so, how can it be explored/implemented in a way that centres ethics, safety, road users, pedestrians and existing social welfare? If not, then should it be implemented at all?