It may be Halloween, but we could do without the hype or horror when it comes to AI in government

It is Halloween once more: our yearly celebration involving costumes, candy and trick-or-treating. If you take a walk tonight in many cities and towns around the world, you’ll likely come across a carved pumpkin on a windowsill or porch, lowly flickering from its yellow grimace; excited screaming children reacting with delight and fear.

More than possibly any other, Halloween is a day that involves both horror and hype. Scary costumes, frightful stories and horror movies come alongside parties and bonfires. These two extremes seem to contradict each other, but just like today, they often go together.

In the current debate surrounding the use of Artificial Intelligence in government we are also, too often, stuck in the horror or hype. While ‘Sophia’, a humanoid animatronic robot, whose creators have been hyping-up expectations by claiming that it somehow is powered by “AI”, was granted its first visa this week, other observers are dreading the looming AI apocalypse.

And while it can be wonderful to indulge in these extremes, when it comes to AI in government, we want to give you the more sober reality. Both hype and horror are fuelled by myths, which cloud our thinking. Below, we address three myths about AI in the public sector that we often hear.


Myth 1: To make AI work in government, we need cutting-edge technology

It is often assumed that in order to use AI in government, we need the latest technology and a team of world-leading experts. In reality, the technology behind most of what we call “AI” has been around for decades.

Just like a carved pumpkin may seem more impressive from afar, from the outside AI can also seem almost magical, like a tool from the future, detached from today’s public service reality. However, when you examine it closer, the magical, flickering light turns out to be a mere candle.

Of course, you can have a nicer pumpkin, a sharper knife and a better design – just like you can use more elaborate algorithms – but any average pumpkin and kitchen knife will already produce impressive results.

It turns out that the latest cutting-edge “AI” technology is often not required to achieve good results. Many useful insights can be gained by using well-established methods without the need for using the latest science.


Myth 2: We can just build an AI system and move on to the next challenge

The ongoing maintenance of AI systems must be taken seriously. We cannot assume that technical teams can simply build a tool and, once it is deployed, move onto the next challenge. Maintaining AI systems, including altering them in response to social change, takes time, resources and effort, which must be properly planned for and considered throughout a system’s life cycle.

Just like anyone who has ever left a pumpkin rotting on their porch will appreciate, any complex system that is left alone for too long will decompose. The environment, be it bacteria and oxygen or changes in laws and societal norms, needs to be taken seriously. In order for AI systems to be able to reflect the ever-changing world we live it, we need to tend to them.


Myth 3: Using AI means putting people out of jobs

AI systems are tools, not workers. They, with few exceptions, won’t replace an entire job. Instead, they will reshape roles by replacing existing tools or by taking on narrow, well-defined tasks. By involving civil servants, including frontline workers, as co-designers of AI systems, we can identify the tasks that AI can help with and free up staff for more creative, problem-solving, people-facing work that improves service delivery and citizen outcomes. A pumpkin isn’t Halloween – it’s just one part of Halloween. Similarly, AI will likely never become the whole public service – but it can be part of it.

If you want to read more about how to take a realistic approach to building AI in government, and how to do so in a way that promotes legitimacy, have a read of our new paper, ‘How to make AI work in government and for people.

Every year, we celebrate Halloween. Dwelling on the holiday’s hypes and horrors offers a nice distraction from our everyday. But AI in government is not confined to one day – it has to work all year round. Collaboratively and as a society, let’s take a realistic approach to it and examine what is doable, desirable and what isn’t. This way we can avoid the unhelpful hype and horror.

The Centre for Public Impact is investigating the way in which artificial intelligence (AI) 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.