Our Communications Manager Tuomas writes about why it’s important to make public service AIs more transparent. The issues have more to do with the ethical use of our personal data than Rise Of The Machines, it seems.
There’s a difference between information that’s available and information that’s available and widely understood.
This is perhaps the most crucial issue when it comes to the increasing use of Artificial Intelligence in public services. And yet, even a basic understanding of what AI is can hardly be called common knowledge. Try reading almost any AI algorithm documentation as a non-engineer and get ready to feel like you’re transcribing the sacred texts of the Klingon Empire with Google Translate.
Sure enough, we all have to live with having to trust authorities and specialists in matters we ourselves may not be that well-versed with. But with public service AI, the tricky part is that we might unknowingly consent to trust the algorithms with some pretty sensitive personal data.
In a democratic system, the lack of such knowledge makes decision-making more and more, well, undemocratic. And while your hometown’s library AI is not going to go all Skynet on you (probably), there are still valid ethical concerns with any public service AIs. The level of seriousness varies greatly depending on what the algorithm in question does, of course. Does it gain access to your full healthcare history or know the fact that you borrowed all Twilight novels five times this year? Your mileage may vary in terms of which example makes you more uneasy.
There are valid ethical concerns with any public service AIs.
AI in public services is not some passing folly, either. For example, almost all tax decisions in Finland are currently made using AI. At this point, any other option would not be viable, due to the huge amount of resources required for the job. Meanwhile, social insurance institution KELA is looking into utilising AI to make their response times faster.
The potential of these large amounts of private data is such that it sometimes takes even the institutions handling it by surprise. In 2017, the city of Espoo used AI to analyse early childhood education and social data. Based on this, they were able to pinpoint 280 risk factors that might lead to child welfare getting involved later on. Helpful in improving early prevention? Absolutely. Potentially problematic and eerily dystopian? You bet your humanoid ass it is.
The point is: in an ideal world, these are the kind of discussions we should be able to partake in. And to do that, we need to make AI less disconnected from the public discourse. That abstract-looking algorithm has real-world consequences.
This increase in transparency is exactly what our customer Saidot seeks to do. Their AI Registers are websites that present AI algorithms in a way that makes it easy for anyone to understand what they do and why. Helsinki and Amsterdam are the first two cities in the world to utilise this idea. Their respective AI Register sites opened to the public in 2020, showcasing public services that use AI in one way or another. When it comes to informing the public about AI, this is a big step forward; historical, even.
The information on the websites starts with the basic stuff: what is this AI for, why is it necessary and what kind of data does it need to work? This, in most cases, is everything most of us will ever need to know to be able to engage in the conversation. But the beauty of the Register is that it’s not some AI For Dummies type of thing. It contains the more technical information too; you just need to dig a little bit deeper to get into that. In Saidot’s model, the more advanced stuff is simply no longer the first or only thing you see, making the whole thing a whole lot less intimidating or incomprehensible.
Instead of a rise of robotic overlords, the current ethical issues come back to some good old-fashioned, inherently human gatekeeping.
AI is very much part of our present right now. Instead of a rise of robotic overlords though, the current ethical issues come back to some good old-fashioned, inherently human gatekeeping. One cannot have a participatory political discussion if large amounts of people do not know what the discussion is – or that there even is a discussion. In order to bridge the gap, we should be paving the way for a world where necessary information is both easily accessed – and easy to understand.