AI Won’t Make Disability Disappear - and That Shouldn’t Be the Goal

I've had Osteogenesis Imperfecta - more commonly known as brittle bone disease - my entire life.

Technology has changed considerably during that time. My disability hasn't.

That's an important distinction when we talk about what artificial intelligence might do for disabled people.

There's a tendency in technology to frame disability as a problem waiting for an engineering solution. Give us enough computing power, enough data and a sufficiently clever model and perhaps we can finally "solve" it.

As an engineer, I understand the temptation.

As a disabled person, I'm considerably less enthusiastic about being considered a bug.

AI has enormous potential to improve the lives of disabled people. But I don't think its goal should be to make disability disappear. It should be to remove some of the barriers that make being disabled unnecessarily difficult.

There's a difference.

I've Never Needed Technology to Make Me "Normal"

Growing up with Osteogenesis Imperfecta means I've spent my life adapting.

There are things I do differently. There are things that take me longer. There are things I simply can't do.

I've also become extremely good at looking at an environment and figuring out how I'm going to make it work for me.

That's something technology can help with.

What I don't particularly need is a machine deciding that the ultimate measure of its success is how closely it can make my life resemble that of somebody without a disability.

The problem with that thinking is that it starts with the assumption that the disabled person is the thing that needs fixing.

Quite often, we aren't.

Sometimes it's the staircase.

Technology Is Most Useful When It Gives Us Options

Some of the most transformative technologies for disabled people aren't particularly dramatic.

Voice control can make a computer accessible to somebody who struggles to use a keyboard. Computer vision can describe an environment to somebody who can't see it. Predictive text can make communication considerably faster for someone with limited movement.

AI can potentially take this much further.

Imagine a wheelchair that understands its environment well enough to help navigate obstacles. A computer interface that adapts itself to someone's changing physical abilities. An assistant capable of understanding what somebody is trying to accomplish and helping them do it without requiring twenty interactions with a touchscreen.

None of those things "cure" disability.

They provide independence.

And independence is a much more interesting engineering problem.

Sometimes the Environment Is the Disability

One thing you learn fairly quickly when you're disabled is how much of the world assumes a particular kind of human being will be using it.

They'll be roughly a certain height. They'll be able to walk. They'll have two hands available. They'll be able to see the screen, hear the announcement and reach the button.

When you're outside those assumptions, things become interesting.

I've encountered plenty of situations where my actual physical condition wasn't particularly limiting, but somebody's design decision was.

That's why I think some of the most exciting applications of AI aren't about changing disabled people at all.

They're about making environments better at adapting to us.

A genuinely intelligent building, for example, shouldn't require everyone to interact with it in exactly the same way. Neither should a car, computer, workplace or public service.

We've spent decades making humans adapt to interfaces.

Perhaps the interfaces could finally do some of the work.

Please Talk to Disabled People First

This should be obvious, but technology has an impressive ability to make obvious things surprisingly difficult.

If you're developing technology for disabled people, involve disabled people.

Not at the end.

Not when the prototype is finished and somebody remembers there was an accessibility section in the requirements document.

At the beginning.

There is a fundamental difference between imagining what would make somebody's life easier and actually living that life.

Engineers are particularly vulnerable to this because we like solving problems. Give us an interesting technical challenge and we'll happily disappear for six months and emerge with an extraordinarily sophisticated solution.

Occasionally somebody then has to point out that nobody wanted it.

I've been on both sides of that equation: I'm an engineer and I'm disabled.

If there's one thing that combination has taught me, it's that the technically impressive solution isn't necessarily the useful one.

AI Doesn't Get to Decide What "Better" Means

There's also a more uncomfortable question underneath all of this.

Who decides what constitutes an improvement?

If AI becomes capable of compensating for more disabilities, that's potentially extraordinary. But the objective should remain with the person using the technology.

For one person, that might mean walking.

For somebody else, it might mean having a better wheelchair.

For another, it could simply mean being able to turn the heating on without asking somebody for help.

Those are all legitimate goals.

The technology industry needs to resist deciding what disabled people ought to want from technology.

Give us tools.

Give us options.

Then let us decide.

I Want Better Technology, Not a Different Me

I'm excited about what AI could do for accessibility.

I'm an engineer working with AI, sensors and intelligent systems, so it would be slightly strange if I weren't.

But I'm much more interested in technology that gives people greater control over their lives than technology designed around some abstract idea of making humans "better".

I've lived with Osteogenesis Imperfecta for my entire life. It has caused difficulties, pain and limitations that I would be lying if I pretended were somehow wonderful.

But it has also shaped my life, my perspective and, almost certainly, the way I approach engineering problems.

I don't need an AI system to have an opinion about whether that makes me better or worse.

I need it to be useful.

Perhaps that's a less spectacular ambition than curing disability.

But giving millions of people more independence, more choice and fewer unnecessary barriers seems pretty ambitious to me.