Do Dutch politicians need a crash course in Data Science?
A month ago, the well known Dutch tv show “Zondag met Lubach” discussed the knowledge of Dutch politicians on digital technology, or rather, the lack thereof. The presenter, Arjen Lubach, spoke of a digital illiterate democracy. While I have always been interested in politics, I had never thought about the influence of data science on Dutch politics and the Dutch government before. With the Dutch elections coming up, it is an important and just as well interesting topic to discuss. Are the government and politicians even aware of the issues and opportunities of digital technology and more specifically, data science? In this article we’ll delve into the current state of data science in Dutch politics, with the help of the expert knowledge of Bert-Jan Butijn (PhD student at Jheronimus Academy of Data Science) and Shirley Kempeneer (Assistant professor at Tilburg University).
Current state
Are things really as bad as how Arjen Lubach portrays them? “Today, there is still little digital expertise in the House of Representatives”, Shirley Kempeneer argues. “That definitely needs to change.” She is Assistant Professor at Tilburg University and teaches the course Governance and policymaking for data science master students. While she does think that there is too little knowledge on data science in the Dutch House of Representatives at the moment, there are some good steps in the right direction. The House of Representatives voted in favor of the appointment of a permanent committee of Digital Affairs after a report (appropriately named “Update required”, they sure do know their puns!) recommended it. She also mentions that the paper headlines do exaggerate the use of algorithms by the government. “The final decision is often still with humans.” The Netherlands Court of Audit recently screened all of the algorithms of the government and found that most of them are still simple and transparent, but that there is a need for more regulation and transparency for civilians.
Another burning question I had was if decisions that are made in the House of Representatives are backed by data or models. Certainly, the evidence (backed by data) or predictions (made by models) that data science can provide, must be enough to objectively pass or refute proposals. It could potentially make the process of legislation much more efficient and transparent. Shirley explains: “The dream of evidence-based policy making has been there for a longer time, but in practice is hard to realize.” She describes that while for a longer time analysts have been doing predictions with smaller qualitative datasets and surveys, policymakers do not always do something with this information. “On the one hand, making policy is almost never purely rational. Policy makers often take into account a lot of aspects and also rely on previous experiences, gut feeling and electoral considerations while making decisions. On the other hand, the statisticians and analysts do not speak the same language as policymakers. They can’t easily convey their results and do not always have an answer for the questions or issues the policymakers have.”
“The dream of evidence-based policy making has been there for a longer time, but in practice is hard to realize.”
Bert-Jan Butijn, currently doing research into smart contracting and blockchain at the JADS in ‘s-Hertogenbosch, but also having been a municipal council member and interested in politics, echoes the same issues: “When presented with the exact same report and numbers, politicians from different parties turn these into diverging stories and come to conclusions that are miles apart. I do not want to speculate that this is the result of unwillingness, ignorance or a bias towards their ideals, but regardless of that, it’s hard to reduce a report of often over 100 pages with a lot of numbers to its essence and still be on the same page.”
Improving the use of data science
Bert-Jan: “Before you can perform any analysis as a data scientist, you’ll need cleaned data to perform an analysis on. The systems and databases of the government are very old in that regard, performing a query in SQL wouldn’t even be possible.” While this wasn’t a surprise for me (I’ve read and heard stories about older systems and messy structures before), it does seem like it is holding back all of the further development of data science in politics and governance. Shirley also recognizes the need for a basic data infrastructure and standardization in data governance. “All public sector organizations and their internal departments have a lot of data, but the quality of it varies widely, is often not or hardly connectable, and there is an absence of the necessary basic infrastructure to capture, store, clean and analyze the data in an efficient and secure way. If people already fear discrimination by algorithms and you would like to reassure them, it is particularly important to train your algorithms on good quality data, which is generally lacking.”
Regulation: A blessing or a curse?
While both Shirley and Bert-Jan argued for the implementation of a basic data infrastructure, it also raises another question: are these developments even favorable and doesn’t the government need to focus on regulation of these technologies first? It is safe to say that a lot of companies in this field are very far in the development of data science and the government is always last in adapting to new technologies. Shirley: “Transparency, accountability and privacy have been more in the political picture for a while now. Despite that, there is still a lot of room for improvement. It’s not always clear which authority is the supervisor.” This has made it possible for companies to do much more with our personal data than we might have wanted to. Shirley put us at ease regarding the use of algorithms (they still were simple and transparent), however, she also mentions the tampering and mistakes with data science in the public sector, resulting from the lack of infrastructure and regulation. For example, the large scandal regarding childcare benefits that made the third Rutte cabinet resign on the 15th of January has been deemed discriminatory and in violation with the GDPR by the Dutch Data Protection Authority. While you could assign this mistake to inexperience with technology, it did have serious negative consequences for a large number of citizens. This case demonstrates that the trade-off between regulation and development is a tricky business. When Bert-Jan was asked about the need for regulation, he answered: “It would be good to have more regulation and the GDPR is a positive step in that direction. However, equally important is the funding for scientific research and innovation. Data is the new oil after all.”
Elections
With the presidential elections in the United States last November and the upcoming Dutch elections in mind, Bert-Jan mentions the difference in operations of political parties during the elections. “Political parties, especially in the United States, use advanced predictive analytics to predict with high accuracy how many people will vote for which party. They not only predict what people will vote, but they can also analyze which messages will persuade people and exactly where their voters are located.” While the Cambridge Analytica data scandal in 2018 showed us the extent and scope of the use of data science during campaigns in the United States (political data science, as Bert-Jan likes to coin the term), in The Netherlands it is still in its infancy. A large part of this has to do with the budgets of the campaigns. In the US, the campaigns of the parties have a large amount at their disposal, while in The Netherlands, political parties do not have large sums of money to spend on their campaigns. They thrive on volunteers, Bert-Jan knows from experience, as they usually cannot afford a full-fledged data science team.
“In The Netherlands, political data science is still in its infancy”
Although there is increasing attention for data science and in particular transparent algorithms and privacy, the subject is still far from being one of the more important topics of the upcoming elections. “A good reason for this is that the subject remains rather abstract and unimportant for the average citizen”, Bert Jan Butijn emphasizes. So what can we, data science enthusiasts, change about this? One of the simplest and most feasible ways, of course, is to cast your vote during elections. Vote for someone with expertise in digital affairs, or vote for a party that describes this as an important point in their election programme. For the upcoming Dutch elections, the Technologie Kieswijzer can help you decide your vote. Furthermore, the data journalism blog Pointer wrote an article summarizing all viewpoints of parties regarding personal data and algorithms.
Do Dutch politicians actually need a crash course in Data Science? Some of the things Bert-Jan and Shirley shared with us could be classified as alarming. For politicians though, its probably not too high on the list of many (long term) issues that they have to deal with. However, I do believe that data science can be a key element in solving all of the other issues. It might be necessary to initiate a mandatory “Perspectives on Data Science” course for politicians, which introduced me to data science three years ago as a freshman. In that course, I had a lecture from Professor Van der Aalst, but I was not yet aware of his status in the field. Had I known, I would’ve probably paid more attention, but in consolation I did find a very relevant quote from him to end this article with: “If Data Science is the new oil, let’s avoid pollution.”
