Jan van Ginkel is deputy provincial secretary in the province of South Holland and a much sought-after speaker on digitalisation. In-house, he is very much engaged in digital transformation. Responsible innovation with AI and algorithms is the new normal, he argues. His answer to how he leads on innovating responsibly in AI and algorithms: “Giving space, investing, having the right people in place and consciously developing ourselves”.
ICT plays an important role in almost everything the province of South Holland does. The deployment of algorithms and AI is part of all of this. Van Ginkel illustrates: “The province is responsible for managing roads and waterways. We deploy AI to assist bridge operators in picking the best time to open a bridge. This way, we can improve the traffic flows for road users and shipping.” Van Ginkel emphasises, “We always apply AI to help move societal challenges a little further ahead.” View South Holland’s algorithms in the Algorithm Register.
Generative AI
South Holland is also deploying AI and algorithms in its day-to-day operations. For instance, PZH-assist, a generative AI application, was developed in-house. “Employees can use it to ask all kinds of ChatGPT-like questions”. It was a conscious decision to develop the solution in-house: “If you offer commercial AI tools a document with the prompt ‘create a summary’, the Googles of this world can read along. Our own version runs in a private environment and is completely secure. In addition, we are currently developing applications that allow our own data to be queried via a chat bot, for instance a chat bot that can answer queries about our leaves and IKB-schemes. PZH-Assist is already attracting considerable attention. A community of people are now looking into translating this to their own organisations and implementing it there. We are happy to share our solutions, after all, it is already paid for with public money.” More on PZH Assist.
New normal
Responsible innovation is a self-explanatory matter to Van Ginkel. “We are a government body, so transparency, reliability and democratic rights are important values for us,” he says. However, he does point out that provinces mainly use geodata. “When processing personal data, privacy and ethics weigh more heavily. We take the moral-ethical side very seriously, but we can develop a lot without personal data.” South Holland safeguards the ethical side in various ways. “A team of ethicists is in constant dialogue with the organisation to help develop awareness, importance and skills. Responsible innovation is the new normal I want to move forward to.”
FRAIA
Of course, that also involves dilemmas. “We do work on that through sessions like FRAIAs.” Data minimisation is an important starting point at the province, Van Ginkel explains. “The constant challenge is: how can you create a good solution with as little data as possible? For example: at what level of detail do you buy geodata? At too high a level of detail, that information might become privacy-sensitive after all. Also, when creating an algorithm to optimise bridge operation for instance, some information about ships and road users will still be needed. However, I don’t need to know whether the skipper’s name is Piet or Jan.”
Discussion tool for politics
Building knowledge in-house
How does Van Ginkel lead on responsible use of algorithms and AI? “Giving space, investing in it, having the right people in place, but also consciously developing ourselves”. The latter was a conscious decision when developing PZH-Assist: “We want our employees to have full understanding of how generative AI works. So, we asked ourselves, can we programme it ourselves? And as it turns out, yes, we can.” Knowledge is important. “If you want to participate in the digital transformation, you can’t outsource or direct everything: you have to build knowledge yourself. That’s what I want to invest in.” Therefore, to Van Ginkel, knowledge of the power of data and algorithms should not be limited to engineers. He says, “Understanding that the world of data and algorithms is now a dominant world that directly affects each and every function of government, and turning that awareness into knowledge and action, is the main challenge for government, I think.”
Scalable solutions
Van Ginkel is proud the province is developing so many things by itself. “In case we cannot do it ourselves, we collaborate with a network of startups and universities. Also, we have teamed up with four other provinces to recruit eight top AI trainees. They will be helping the five of us in moving several societal issues ahead and ensuring that those solutions are scalable (to be used by others, ed.)” This remains a tricky issue, he says: “If we create something with AI around housing and infrastructure, it is often not immediately useful to others. The concept of a house in Flevoland for instance may very well differ from ours.” He now also aims to computerise that translation. “I try to make scaling up part of the algorithmic solution,” he says.
Principally relevant
The province of South Holland was one of the forerunners in the creation of the Algorithm Register. Van Ginkel therefore fully supports the register but also places a critical comment. “I said earlier: the government has to be reliable. Therefore, it is only natural that the government is transparent about the algorithms it deploys. For me, this is a principally relevant matter that must be put into action. An algorithm register is one of the ways to do that.” The organisation benefits, he says. “We refer citizens, companies and institutions to it to show people how we do things.”
However, he comments: “The number of algorithms is increasing exponentially: 10 are added today, 100 tomorrow and 1.000 the day after tomorrow. The registry is largely being filled out manually. So, my point of critique is: how sustainable is an algorithm register like this? I would understand if you said the register is meant mainly to keep organisations aware and transparent to citizens. But if you want the register to be complete, in the long run maintaining it in that way is just not feasible. These days, when you buy an accounting package, it includes algorithms. I personally don’t have a solution for this yet.” He concludes: “I support the algorithm register wholeheartedly, but there is an element of naivety to it.”
Innovating together
So, is there a word of advice for other administrators? “My appeal is: let us work interdepartmentally and intergovernmentally with far more pace and vigour. Innovate together, to achieve results in societal issues. For instance, the issue of grid congestion (overloading the energy network, ed.) affects the state, it affects municipalities, it affects water boards, and it affects markets … How about if we all said: this is where we will set up not eight, but eighty AI trainees?”