Artificial Intelligence and Disability
As the deadline for my major project inches ever closer I find myself scrambling to cover blind spots. The main question that I have sought to answer in this project how does AI affects regular everyday people. Intellectual arguments aside, new technologies change our lives, for better or for worse. Ethical boundaries are constantly being pushed by technology and I want to get a grasp of where and how AI redefining what it means to live in our world. One of the inherent problems with a broad topic such as this is looking beyond your own experience. As an able-bodied person I tend to forget that the perspectives of those living and working with disabilities needs to be part of this conversation.
To this end I read through an excellent article by Francisco Jose Bariffi entitled Artificial Intelligence, Human Rights and Disability. In it Bariffi identifies several key areas of concern with regards to AI and its impact on those living with disabilities.
The “Reverse Turing Test”
As Bariffi points out AI systems are often designed and implemented by the able-bodied. He notes that many voice recognition systems struggle to interpret the speech patterns of those with disabilities, such as down syndrome, resulting in a perverse “reverse Turing test” in which individuals need to prove their humanity to a machine. This is due to the programmers and designers making the assumption that everyone speaks like they do. As Bariffi points out the failure to accommodate someone with a disability is a form of discrimination regardless the intentions of those who designed the AI system (harm is harm whether you intend it or not). These types of omissions contribute to the invisibility of disability in our society where services are designed only for the able bodied and all others are an afterthought. This can be seen in the accessibility of our buildings, businesses, services, parks, recreation, and even on the internet itself.
Employment and Inclusion in the Workplace
Bariffi argues that inclusion of people with disabilities in society hinges on their ability to access education and meaningful employment. However, this is being put in significant jeopardy by the AI revolution. The push to employ AI based decision in human resources by large employers has opened up opportunities for companies to screen out those with disabilities who are applying for positions. While this is highly illegal in most jurisdictions he argues the opacity of AI algorithms (many companies consider their proprietary algorithms trade secretes and closely guard their inner workings) makes them ripe for abuse. Without some sort of framework for making companies “open their books” to regulators it is possible that the gains made those with disabilities over the last several decades could be slowly eroded. As an educator this possibility is particularly repugnant to me as the heart of our current inclusion strategies could be rendered moot to for our graduates.
Some Light Amongst the Darkness
It should be noted that AI provides many opportunities for those living with disabilities as well. As Bariffi points out one of the fundamental strengths of AI is that it allows for it to quickly adapt to changing circumstances. This could take the form of customized assistance for those who have solely relied on caregivers up to this point. As Bariffi argues independence and self sufficiency may be facilitated by eliminating the need for personal care assistants in some situations. He gives the example of the incredible abilities of AI software to accurately describe pictures to those with visual impairments. This is a far cry from early text to speech systems with their robotic and stilted delivery.
The Implications of Deepfakes
In a previous blog post we discussed how most funding for AI research is coming from venture capitalists who are overwhelming white and male. This got me thinking about how AI will specifically impact women. While searching databases I stumbled on an article by Chidera Okolie exploring the dark side of AI-powered deepfake technology.
As she points out deepfakes use AI-based technology to face-swap one person for another in a still image or video. A far cry from the early days of photoshop these fabrications are extremely convincing and have been used for a myriad of legitimate and nefarious purposes.
From making politicians say things that they did not say, to hoaxes, forgery, and extortion the ability to easily make anyone do anything is a terrifying and shockingly easy prospect. As she points out anyone with time and basic graphic design skills can create facsimiles with very little effort. This poses threats to democracy and the basic dissemination of information as it becomes difficult to tell forgery from truth. Disturbingly she goes on to detail how the pornography industry has co-opted these techniques to create sexually explicit material without any regard for consent (motivations vary, but common ones include profit, revenge, pubic shaming, and ‘sextortion’). The implications of this use case are terrifying. Imagine the pain, humiliation, and professional repercussions of someone using your likeness in explicit material against your will. As Okolie notes this disproportionately affects women, rather than men.
In my next post I am going to further investigate the abuses around privacy and hopefully begin to start synthesizing my thoughts.