The ongoing conversation and development around artificial intelligence are bringing data policy back into the spotlight, as we will explore in a dedicated session at the e-Governance Conference 2023. Perhaps it’s also a signal, an indicator of how much work and care is needed before AI starts convincingly scaling to the public sector.
Because while it is true that technological advancement can unlock new levels of digital governance, the great reliance of AI models on enormous amounts of information refocuses the importance of data quality, management practices, ways and purposes of use.
For an introduction to data policy and artificial intelligence in government, we are joined by Paul Timmers, Research Fellow at the University of Oxford and Senior Advisor, and Heiko Vainsalu, Programme Director on Technology. Plans, expectations, democratic legitimacy – how disruptive is AI set to become, exactly, in and to public sectors worldwide?
Rethinking frameworks for data policy around citizens
First things first. Data policy and data governance – what is it all about? Without venturing into formal definitions, with these terms “we are talking about the way you organize yourself to deal with data, and the rules and policies around it,” Timmers begins with.
According to him, on a more practical level, three levels or actors shape up the questions that feed the discussion around data policy:
With these three categories, we’ve come to represent society as a whole. And to the end of delivering public services digitally, we can see how the topic has far-reaching implications.
“If I put some data out in the field, what kind of services should I expect in return? Whether these come from a government or another service provider. Governments find it easier to be on the good side, saying that your data will remain private, that we’ll use it only when there is a legitimate interest, that we’ll follow security principles, and so on. But at the same time, when data goes into the private sector, we do see very cool services – but hardly or very rarely is discussed what kind of principles are followed there,” Vainsalu adds.
Who is talking about data and governance today?
It is interesting to note, already, that the conversation over data use and management has pretty much turned into a public debate. Not something to take for granted – think of 15 years ago, for example. “When I started my career in the public sector, data was very specific and strongly controlled by very strong players. And discussions over data governance were held among a very limited set of stakeholders in Estonia,” Vainsalu recalls.
“Today, there is so much data and many more stakeholders are involved. Data governance is a topic addressing a wider scenery than ever before, calling in a wider audience to participate so that all stakeholders involved can understand what is going to be done with certain data, and what can be expected from others with regard to their use.”
If we were to go out on a street and ask bypassers what good data policy looks like, they probably wouldn’t have an answer. “But if you ask whether it should be private, or where it should be stored, some people will already have an idea about the possible answers. It’s not anymore a discussion just for professors or IT experts,” Vainsalu explains.
With public discussion should come awareness. Something that corporate executives have very well in mind already. “In a recent survey in the Netherlands, about 20 CIOs have been asked what’s the most salient element to consider in digital transformation. And the most popular answer was not skills or cybersecurity – it was data. And that’s what they see their business revolving around,” Timmers points out. “I’m not saying public administrations should behave identically to companies. But this awareness, that data really is the key asset of your organization, seems to be just emerging in public administrations today.”
“I can only say it would be irresponsible of governments to not address this topic, the importance of data, and especially of data that relate to citizens. What’s at stake is not just that, but the overall credibility of the government itself. It’s not easy to do a good job at it, but there are countries that have shown this is possible,” Timmers says.
Artificial intelligence needs data – but to do what and how?
On a larger scale, the increasing applications of artificial intelligence models seem to be forcing governments to rethink the role of data and the ways they manage them for public service provision. But it’s hard to predict what will AI in public administration look like, “simply because we haven’t lived in such society yet,” Vainsalu points out.
“Artificial intelligence is something that thrives on enormous amounts of data, coming from a lot of different sources. To understand what could happen, we need to see when in the past a resource [such as data] has become so abundant, and how that changed the way society, the economy, or industry work. We are in a similar paradigm to what happened with the advent of digital photography – quantity is thriving, but quality is falling behind. We are yet to see how the massive use and abuse of this new, abundant resource will start to reshape the world,” Vainsalu explains.
One thing we know though – while learning about algorithms and how to effectively use them, we must keep a critical approach. For example towards classifiers, which Timmers recently wrote about with scholar Paul Waller. “Any algorithmic process used by public authorities to predict attributes or circumstances of people”, say the authors of the paper, “should not happen in isolation, not be a single factor for decision-making.”
“People are essentially reduced to single data points. Are you someone who is likely to commit fraud on social benefits? Or are you deserving of a visa or not? If you are judged on the basis of data that has been simplistically collected through a classifier algorithm, holding a critical stance is the least,” Timmers adds.
Resisting the temptation to chase the private sector
As happened before in the case of other technology-turned buzzwords, the private sector is showing plenty of interest and activity to bring AI to markets. A legitimate fear is that public sectors worldwide might feel the rush to roll out AI-powered services to chase after companies’ product development.
“This is actually the thing I’m very much afraid of, that the private sector will show new services with an impact that will build expectations. Expectations towards the public sector to do the same, but come to life in a much less regulated environment. AI could be used in the public sector, but not because of the influence of the private. Rather, because people who are designing services see what kind of value to deliver to people, are able to understand benefits and risks, and are able to consider them properly,” Vainsalu says.
Data policy and AI, ultimately, so become the playing field where governments can even go beyond just a careful approach. “I would challenge the public sector to take more of a leadership role. We want to have ethical AI, and responsible data governance, and this would just be in the very nature of ministries and public administrations. Subjecting use and purposes to public discussion, explain what you do with the data, show the algorithms, and make the process transparent,” Timmers concludes in his call to action.
Interested in rethinking frameworks for data policy?
Join Paul Timmers, Heiko Vainsalu, Ott Velsberg – Government Chief Data Officer of Estonia, Claudia Oliveira – Programme Manager at EU Commission’s Department for Informatics (DG DIGIT), and Amos Mpungu – Principal ICT Officer, Ministry of ICT and National Guidance, Uganda in the discussions “Data deluge – do we control data or it controls us?” at the e-Governance Conference on 31 May at 11:00 – 12:30. And whether in person or online, join us to build better and inclusive digital societies!
