The Netherlands Authority for Consumers and Markets (ACM) works with a wide range of algorithms. Thanks in part to the national Algorithm Register, the organization gained more insight into which algorithms it uses and for what purpose.
There are currently 15 ACM algorithms in theAlgorithm Register. This makes the authority a pioneer among regulators in the Netherlands. But that is only the beginning. This number is only going to increase in the coming period, according to board member Martijn Ridderbos and Kari Spijker, algorithm governance specialist at ACM's Task Force on Data and Algorithms (TDA).
"We have been working on the national Algorithm Register for almost two years now," says Spijker. "It has helped us to get a better grasp of all the algorithms we have in-house. By registering our algorithms in the Algorithm Register, we were forced to think about the question: what do we actually define as an algorithm, and what does and does not fall under that definition?"
According to Spijker, the register therefore helps to create an overview. “The ACM consists of many different departments, each of which supervises in its own way. The Algorithm Register prompted us to update and expand our internal register. Which algorithms and AI systems do we use? And are they in order? The national Algorithm Register has really helped us to work even more responsibly and transparently as a supervisory authority.”
Ridderbos endorses that view. "Our mission is to ensure that markets work well for people and businesses, now and in the future. Based on that mission, we have three core values: independence, openness, and resourcefulness. It is precisely because of that openness that the Algorithm Register is so important. We believe it is essential to be public and transparent about what we do. That is why we share our most important algorithms in a register that is accessible to everyone."
However important transparency may be, ACM cannot disclose all algorithms in full because that would jeopardize its role as a regulator. “You don’t want everyone to know exactly how some algorithms work,” says Ridderbos. “That could undermine supervision. It’s a dilemma, because we want to be as open as possible.”
That is why some algorithms are only described in broad terms. "We rephrase information that is too sensitive," explains Spijker. "For example, we describe in general terms how an algorithm works technically, without revealing details about sources or specific methods."
Ridderbos emphasizes that this is not a random choice. "We do not withhold information that the public should know based on the criteria. But we do carefully consider how to prevent our supervisory tasks from being compromised."
The Data and Algorithms Task Force is also increasingly developing its own algorithms that provide greater insight into markets and market behavior and enable the detection of market problems. "We expect that more registrations in the register will follow in this area," says Ridderbos. "Developments are happening at lightning speed, both at the parties we supervise and at our own organization. We are looking at how we can use AI to enrich our work and are also using that knowledge to effectively supervise markets in which data and AI are playing an increasingly important role. This requires maximum transparency where possible and due care. The more we work with AI and algorithms in the Netherlands, the greater the need to be open about it. This is certainly true for public service providers. We hope that other parties will follow suit. This will contribute to a level playing field.”
Due to the increasing use of data and algorithms, ACM established the Data and Algorithms Task Force in early 2023. “We are currently transforming it into a fully-fledged directorate,” says Ridderbos. “Data-based supervision is an essential part of our task.” Spijker also sees a broader social responsibility in this. “Digitization is radically changing society, and our digital tools and capabilities must grow along with it. Regulators in particular have a responsibility to steer these developments in the right direction and ensure a safe and reliable online environment.” That is why ACM, together with other regulators, the Dutch Healthcare Authority (NZA) and the Netherlands Authority for the Financial Markets (AFM), is training data specialists who specialize in the use of data and algorithms in supervisory practice through a traineeship.
Ridderbos and Spijker are also calling on other government organizations to share their algorithms via the Algorithm Register: "Don't get bogged down in definitions," advises Spijker. "Just start by taking stock and sharing. The Algorithm Register is not only an obligation, but also a source of inspiration. It shows what you are already doing. And you can be proud of that."
