Skip to content
Article Article September 18th, 2017
Technology

Shining a light on AI in government

Article highlights


Magic 'is a good reference point' for thinking about AI, says @aaronsnow

Share article

Governments need to realise that #AI's impact will uproot previous approaches and positions

Share article

Algorithms 'are not ok' for decisions that materially affect people’s lives, says @aaronsnow

Share article

Partnering for Learning

We put our vision for government into practice through learning partner projects that align with our values and help reimagine government so that it works for everyone.

Partner with us

Magic. We've all read the Harry Potter books and seen the films. Wands, the Sorting Hat, the Pensieve - magical things to empower their wielders with knowledge and abilities beyond the mortal. In J.K. Rowling's stories, the wizards and witches are acutely aware of how frightening magic can be to us non-magical “Muggles”, so they do their best to hide all traces of their magical world from us - and even employ an entire government agency, the “Ministry of Magic”, to enforce that invisibility.

Technologies that can appear nearly as supernatural are on their way to our real world, with no Ministry of Magic to govern them. Like magic, artificial intelligence (AI) is poised to dazzle and shock us, but its influence will be felt far beyond the reader's imagination or film studio. From driverless cars to transformed healthcare and justice systems, this technology is not just flights of fancy or sleights of hand - it's real, with real power and consequences.

AI's potential to transform the world is something that Aaron Snow is regarding with a mix of excitement and awe. The founding deputy commissioner of the Technology Transformation Service (TTS) in the US, as well as the cofounder and former executive director of Washington-based 18F, the 200-person digital agency at the heart of the TTS, Snow knows more than most about how to deploy digital technology in government. And he, too, is drawn to the parallels between this fast-evolving technology and the supernatural.

“Arthur C. Clarke once said that ‘any sufficiently advanced technology is indistinguishable from magic',” he recalls. “This is where AI hits. When computers can analyse in seconds what no human or group of humans can hope to achieve in a lifetime, the computer is doing things that no layperson can hope to understand. So ‘magic' feels like a good reference point for thinking about how to regulate and interact with AI.”

Questions in abundance

The sheer pace of the advances of all things AI in recent years has left even its strongest supporters surprised. It is no longer just an abstract theory or the stuff of a Hollywood blockbuster - it's here in real life. With that in mind, perhaps it's not surprising that governments - never the swiftest of organisations - are having to adjust to this fast-shifting environment. And with its role in respect of these technologies yet to be defined, Snow agrees that there is much to clarify.

“Do we regulate it?” he asks. “Will it get done well or poorly, cheaply or expensively? Will it be done quickly or slowly, equitably or geared towards those with power and money? How do we react as a society, government or private sector organisation? Given that we can't all understand everything about how these systems work, what can we hope to learn and understand, and what do we need to understand, in order to interact with this stuff without spiralling into some dystopian sci-fi situation where the AI is making decisions and imposing real-life consequences without our being able to argue the point?”

Such questions need answering, if only to counter the understandable scepticism which many people feel when confronted with the reality of an AI-shaped future. It is also worth considering that if we think of AI as magic, then will we ever trust it? After all, as humans we are not conditioned to think of AI as a series of processes giving us an outcome. This might act as an inhibitor for us to utilise these technologies and take advantage of their massive potential benefits.

“For some people, the reaction will be - and will remain - that fear, that scepticism or rejection,” concedes Snow. “What's happening in technology with things like AI is a more real type of magic than a stage magician's sleight of hand. With stage magicians, we are conditioned to know there is something clever but simple going on - there's a hole in the floor, an extra ace up the sleeve. But AI is much closer to the real deal - there is no sleight of hand about a computer making billions of computations per second. It's far more powerful than a trick,” and thus far more daunting to the uninitiated.

Adjusting to the new reality

Snow believes that as part of their slow progress towards embracing AI, governments need realise that its impact will uproot previous approaches and positions. “AI is coming to a government near you,” he points out. “It is very inherently political, and at some level it is about transparency. We have advocated for many years for open government. But what does it mean to be ‘open' where, even if the algorithms driving decision analysis through AI were made open, only a tiny fraction of the public could understand it - or where machine learning brings machines to a point where even if it was open, no one could understand it?”

That said, he goes on to agree with the suggestion that we need to find a way for the wider population to understand AI and how it reaches its outcomes - akin to an audience being taken behind the curtain of a magic show and shown how the tricks work.

“This is absolutely critical,” he says. “This isn't entertainment. These are decisions about who receives welfare benefits and who goes to jail for longer or for less time, and so on. These are serious decisions that affect people's lives in every conceivable way. Remember, we know from history that decisions being made in back rooms or closed environments is how trust erodes and is inevitably a cause of political upheaval.”

Shining a light on how AI works will reassure citizens about its benign intentions, as will ensuring that systems are put in place to prevent an algorithm from exceeding its original design. “I see an algorithm as an automated decision recommendation system,” says Snow. “What I think most people are fearful of - and rightly so - is putting not just analysis and recommendations in the hands of these systems but final decision-making authority as well. Algorithms are in some ways a perfect mechanism for bureaucracy and operationalising mundane, harmless tasks. But they are not OK for decisions that materially affect people's lives. The higher the stakes, the more careful we need to be and the more humanity needs to be injected into the process.”

From the outside in

Snow is currently working in the private sector as the chief operating officer of cBrain North America, which helps governments digitise their bureaucracies. It seems an ideal role for him to deploy the skills and insights gained from years at the coalface working on technological transformation in the public sector, which some might say is a Herculean undertaking. Asked to identify why government seems slow to move when it comes to technology, he says a number of factors are at play.

“One, there is no alignment in team incentives in government like there is in the private sector,” he points out. “There's no profit-sharing. And there is a lot less control in government over how the business is run, because much of the ‘how' is predetermined in legislation that dictates from afar what agencies can do. And finally, government agencies don't have a lot of control over their organisational goals. Companies, if they are failing, can pivot to change their products and services, whereas government bureaucracies are promoting goals that are not of their own choosing.”

Admittedly, this seems a somewhat bleak picture of day-to-day life in government. Snow, however, is quick to add that governments should definitely seek to take advantage of AI. “AI is going to happen,” he stresses. “Government now has to figure out how it is going to participate in this new world. It won't get better at building and buying digital systems unless it has digital systems experts in-house, and some of them should be doing the building and buying just to make sure they are keeping up.”

There's more. With governments having no choice but to seek to capitalise on AI, Snow believes that its workforce all need to get up to speed - and fast. “They need to learn about how AI is already being used in their area of subject-matter expertise,” he says. “Being aware of what's happening in private industry and other governments is important, as is helping make sure the public is supplied with easily digestible materials to understand this better.”

And as for the rest of us? Snow believes it is incumbent on us all to keep our eyes wide open. “Be vigilant,” he says. “Keep an eye out for decisions that are being made about you and about your life that don't seem to be sanity-checked.” He concludes, however, with a note of reassurance which better reflects his belief that AI is something to be embraced, not feared. “There will always be a very human component of helping people interact with these systems when it comes to government services,” he says.

And perhaps that's the ideal situation for whatever awaits us down the road: AI and humans working together, delivering together. Should that scenario come to pass, we all stand to benefit.

 

The Centre for Public Impact is investigating the way in which artificial intelligence can improve outcomes for citizens.

Are you working in government and interested in how AI applies to your practice? Or are you are an AI practitioner who thinks your tools can have an application in government? If so, please get in touch.

FURTHER READING

Written by:

Joel Tito Fellow, Centre for Public Impact
View biography
Share this article: