For better or worse, the term ‘artificial intelligence’, or simply AI, still conjures up science-fiction-like images of dangerously powerful computers or malevolent robots overthrowing the human race. It still seems like something that belongs in the future, even though it’s been around for a long time and is being used by countless numbers of people on a daily basis, many without even realising.
Online retail websites, social media platforms, film and music streaming services, email filtering systems and ‘virtual assistants’ like Siri and Alexa all use AI, to name just a few uses. And as many e-Access Bulletin readers will know, those virtual assistants also have a wide range of benefits for people with disabilities. So, how can artificial intelligence be harnessed to provide as much assistance as possible for people with impairments? And how can the technology be developed in the future to be of even more use in this are?
These were two of the topics being discussed at an event in Vienna organised by the European Disability Forum (EDF), titled ‘Using artificial intelligence to enhance accessibility – opportunities and risks of emerging technologies for persons with disabilities’.
Specialists from major technology companies were present to talk about products and services using AI, alongside various field experts from disability and AI-focused organisations.
Moderated by Michael Fembek from the Zero Project (an Austrian organisation which shares and supports innovative disability-focused projects), the first session, titled ‘Setting the scene’, featured representatives from three tech giants. As Fembek noted in his introduction, “the relationship of disability, accessibility and inclusion to AI is not straightforward. There are incredible opportunities … but there are also dangers – the danger of widening the digital gap instead of closing it.”
The first speaker was Adina Braha-Honciuc, Government Affairs Manager for accessibility and sustainability at Microsoft, who began by addressing a question that would be returned to throughout the event: what is AI? “It is about computers understanding the world,” was her core definition. “It’s not about ‘human versus machine’,” she added later, perhaps trying to banish those sci-fi-like images.
Braha-Honciuc claimed that “AI makes diversity even more important than it already is,” citing the need for diverse teams to work with the technology for it to reflect diverse thinking.
“People with disabilities need to have a seat at the table in the development of AI, to make sure that AI systems are not biased and are reflective of our entire society,” she said, highlighting what would become a key message of the conference.
Anna-Verena Naether, a public policy senior analyst at Google, defined AI differently: “It’s the science of making machines smart, or making them appear smart.”
AI is crucial to a lot of Google products, Naether explained, including search technology, Google Translate, Gmail and Google Photos. She then discussed Google Lookout, an upcoming mobile app that uses machine learning to help people with sight loss understand their environment. The app captures images of a scene and detects items, people, etc, in the image, before ‘judging’ which are the most relevant items in the photo and speaking these items aloud to the user.
There are still challenges to overcome with the app, said Naehther, including issues with image quality and the potential for misrecognition. She went on to discuss Google’s work on voice access systems, which allow users to control items purely by voice. This technology could be used to control wheelchairs or tilt beds for people with mobility issues, she said.
Monica Desai, Facebook’s Director of Global Public Policy for Connectivity, went a step further, stating “We strongly believe that artificial intelligence is the future of improving accessible experiences at scale.”
She cited Facebook’s automatic alt text and facial recognition tools as examples of how the company is leveraging AI in accessible technology. Both features make Facebook easier to use for people with sight loss and were developed through extensive user-testing with screen-reader users, Desai said.
She ended by talking about captioning videos on Facebook, a much-discussed accessibility issue. Users are not obliged to add captions to their videos, but the company has added features to make the process easier, in the hope of encouraging more captions, Desai said, including “investing in real-time captioning capability” for Facebook Live.
In a separate session on ‘Setting the scene’, Hector Minto, Microsoft’s Accessibility Evangelist, talked about new developments from Microsoft, including a potential Hearing AI app, a version of the popular Seeing AI, which recognises text, objects and faces to help visually impaired users. Currently undergoing research, Hearing AI would learn to recognise sounds to assist people with hearing impairments – such as a doorbell, alarm and specific voices – and describe those sounds to the user.
Minto also flagged up a new accessibility checker for Office 365 which gives live accessibility feedback as someone works on a document.
Following on, Wilfried Kainz from the Zero Project highlighted the Inclov app as an innovative example of AI being used for social good. Inclov is a matchmaking app for people in India with disabilities, using machine learning to create potential matches.
Kainz also highlighted an important point around practical access to these technologies: “We believe that AI is definitely a driving force for assistive technologies, but how can customers and services come together?” One positive example is Tech Able, he said, an assistive technology showroom in Singapore that anyone can visit to test equipment (in soundproof and lightproof rooms, if required), to see what suits their needs before purchasing.
Over the course of the day, other AI projects were showcased and discussed (including a demonstration of the Amazon Echo voice-operated virtual assistant) and deeper issues were raised. Klaus Hockner from the European Commission’s High-Level Expert Group on Artificial Intelligence, stated that one of the key issues around AI was a lack of early education on the subject. “AI skills are not taught at schools or kindergartens – it needs to be implemented from the beginning.”
Later on, Carine Marzin – a member of the EDF’s ICT Expert Group – previewed findings from an upcoming EDF Report on emerging technologies. The report “aims to support the disability movement in engaging with industry and policy makers to ensure that emerging technologies are inclusive,” Marzin said.
The report explores benefits of these technologies for persons with disabilities, concerns around the technologies, and practical recommendations aimed at ensuring inclusivity.
Marzin then discussed risks and concerns from respondents to the report survey. The biggest concern (from 88% of people) was accessibility of emerging technologies, followed by concerns about lack of standardisation, interoperability with current assistive technologies, discrimination and security, among other areas.
The findings demonstrated that despite the many applications – current and potential – that artificial intelligence and other emerging technologies offer to persons with disabilities, there are still serious real-world concerns around their implementation, from those that could benefit the most from their use.
Education on the topic will help address some concerns, but it appears the most crucial element by far relates to the message consistently repeated throughout the event: persons with disabilities need to be involved with this technology at every stage, from initial discussions to design and development. Only then, it seems, does AI stand a fighting chance of being truly inclusive and truly useful for persons with disabilities on a widespread and long-term scale.
Read more about the European Disability Forum’s work at the EDF website.