by Tara L. Campbell

The first time Google popped up a suggested response in my email, a sense of relief hit me—I didn’t have to struggle to find the right thing to say; I could select a response from a few pre-formed sentences. A few months later, similar automated options appeared in my text messages. Suddenly I was given a much-needed tool to deal with the complexities of communication. I could quickly respond without worrying if what I said was odd or wrong.

Not long after, the message options began to evolve. They started to sound like me. The automated system was learning my personal lexicon, and adapting itself to suit my needs. Finally, I had a simple, cost-effective, and intuitive way of working with my communication disorder. Artificial intelligence (AI) is proving itself an evermore effective solution in many settings, but as far as assistive technology goes, it could be vastly improved if we shifted our focus to addressing actual needs.

What is assistive technology?

Assistive technology is what we use to bridge the gaps in a world where only the dominant set of needs are met, giving people with disabilities more opportunities to function as close to independently as possible. Devices such as hearing aids, prosthetics, and wheelchairs are all examples of assistive technology. But computing technologies have widened the array of options so that smartphones, tablets, and the applications that run on them have the potential to act as assistive technology, too. Adding a layer of AI to assistive technology, specifically the remarkable capabilities of machine learning, could provide a different level of assistance that is currently lacking today.

Unfortunately, this unique possibility of advancement is lost in the mad rush of a venture capitalist-driven tech industry. Our technology focuses either on optimizing for convenience for the abled, or at its worst, for frivolity. But mostly, it’s centered on profitability for eager investors. The World Health Organization (WHO) reported as of May 2018 that only 1 in 10 people have access to assistive technology. Even though the types of devices cited in the report were conventional (eyeglasses, wheelchairs, etc.), one of the main barriers to access is cost. Only wealthy countries have access to assistive technology, and of those, the coverage is still lacking. As of 2012 the U.S. Census Bureau reports that nearly 1 in 5 people in the U.S. has a disability, and yet according to an article in Pew Research, disabled Americans are less likely to use technology in the first place. Cost and accessibility are cited as the predominant reasons behind this gap. Lee Huffman, the editor for AccessWorld — a magazine published on the American Foundation for the Blind platform that provides technology news for the blind and visually impaired—was interviewed for an article in U.S. News discussing the costs of assistive technology. Software that is priced at hundreds of dollars, and devices in the thousands, are a requirement for Huffman to actively participate in society. The burden of cost is placed on people with disabilities, an already impoverished demographic, which means the margin of profit for investors is extremely narrow, if not nonexistent altogether. So it’s easy to see why investors are not clamoring over each other to fund assistive technology.

The sophistication of AI is exciting. Advancements are growing exponentially by the month (Moore’s Law be damned) and yet AI is only as good as the datasets they’re based upon, which currently are deeply flawed with racist and sexist results. If AI is unable to account for this level of variance in humanity, then it’s not a stretch to say that AI is currently incapable of accommodating disability differences. Jutta Treviranus, the director of the Inclusive Design Research Centre in Canada states as much in a podcast on CBC Radio Spark. The push in AI design today is to make life a little easier for those who already fit into society’s abled-centric norms.

Accessibility-driven development

Imagine, however, if we started the development of AI by addressing the needs of people with disabilities first.

Communication is one of the logical starting points. An AI could study key cultural and language markers of a population, and then through machine learning, develop a breadth of understanding far wider than the average human brain is able to learn. With this ability, AI could potentially play the role of interpreter for those who cannot read body language, or understand important nuances like tone or inflection. Having AI handle the complicated processing of calling and organizing thoughts around a problem alleviates a burden that affects many people with disabilities. At a basic level, having an AI-enabled device that explains a need or action on one’s behalf eliminates the frustrating, and sometimes dangerous, confusion involved with miscommunications.

The security of knowing there is always someone, or something, available to help could free many people who are trapped alone in the confines of their disability, and empower them to lead more independent lives.

Some of the smallest everyday tasks that abled people take for granted are the difference between a person living relatively independently or living in deep restriction. For example, AI can solve problems such as using rapid response systems during an emergency, or maintaining accurate medical information including medicine dosages and frequency. Human caregivers are currently the only option for a lot of disabled people, but many areas of need can be addressed by AI-enabled assistive technology, also known as Intelligent Assistive Technology (IAT). A multi-national and discipline research article in 2017 examined the current usage of IAT with dementia care, primarily in the U.S. and Switzerland. What the research found is that while there are in-roads now to using AI for people with disabilities, the substantial efficacy gap is in large part due to a lack of focus on the needs of the disability from the start of development. Backwards engineering what works for non-disabled people to work for disabled people does not meet the needs the technology promises.

Everyone is using technology today that overlaps to make life a lot easier for people with disabilities. For example, the means to set up reminders and schedule appointments, to place orders and even learn online, has made the world so much more accessible. Designing AI technology from the perspective of filling the types of needs that keep people with disabilities from fully participating in life would make it even better.

Tara L. Campbell is a speculative fiction and creative nonfiction science writer with a professional background in computer science. She enjoys writing at the intersection of science, technology, and disability. @taraWritesSci |