How Jim Stolze tracks down bias in algorithms with his new company, de Keuringsdienst van Data

Tech entrepreneur Jim Stolze recently sold his company Aigency and started a new project, De Keuringsdienst van Data. With de Keuringsdienst van Data, Stolze fights against errors in algorithms for, among others, the Ministry of Health, Welfare and Sport (VWS). In addition, the ethical issue of artificial intelligence plays a major role in his business operations.

There are many different ideas of what ai is, but what is artificial intelligence according to you?

“Artificial Intelligence is kind of like a quest, a quest to find new possibilities and techniques to make machines smarter and smarter.”

What is the reason you sold Aigency?

 “In 2016 I started Aigency with two partners because we saw that the world was eager for artificial intelligence. Everyone wanted to work with AI, and we were wanted to play a part in that transition. Companies want to achieve more with their data, and we assisted them in that. Heineken was our first customer, a dream customer to start with of course. Heineken had a problem that they could not solve themselves. They purchase raw materials such as bottles worldwide, and this data was stored and analyzed in Excel files. Ultimately, we developed software that should run on any laptop in the Heineken universe. This made their work process clearer and more efficient.”

“However, the market has changed very quickly in recent years. At the time, everything that Aigency did was new, but now you see a kind of professionalization in the market that the clients want to carry out this work themselves. In recent years we have been helping our customers more to design their own AI experiences, leaving only the software engineering for us. At the time, we worked together with Info Support, a software company. Building software suited them better, so it was decided together that they would take over all AI activities. Info Support sees itself as the Ajax of IT, and with the takeover of Aigency they have a Jong-Ajax under the button where they can develop a lot of IT talent. This was also why I and my partners started Aigency, a perfect transition.”

Jim Stolze (third from the left) during the takeover of Aigency. Photo: Norbert Waalboer

How did you come up with the idea to start De Keuringsdienst van Data?

 “When I noticed that AI had entered that new mature phase, a new problem arose. The whole AI world went too fast and steps in development were skipped. With de Keuringsdienst van Data I take one step back to be able to take two more steps forward. By this, I mean that the basis of AI is still data, and I noticed that several of my customers, such as the Ministry of Health, Welfare and Sport (VWS), asked the question: “How do we know whether the data on which we train is good enough and is this data usable or is there bias?” These are questions that an AI engineer cannot answer. This data from the Ministry of VWS had to be handled with great care, but as a result, I learned many new things that I now also offer with de Keuringsdienst van Data.”

“Before we start working on an issue, we also ask ourselves: “What can go wrong here?” Of course, the General Data Protection Regulation (GDPR) must always be considered, so a lot can already be seen in advance. However, this entails something, because you cannot use all data such as gender, age, and ethnic background. Six months later we presented our model with the results. The Ministry of VWS thought it was an impressive model, but they did not know whether you discriminate against a certain age group. We did not know this either, because we did not have this data available. This was part of the process, and with various processing agreements, we eventually gained access to this data, which allowed us to develop a new model. These two models could be placed side by side, so that you could see whether certain population groups, for example, are being discriminated against.”

“Then I thought that we have developed a model that allows us to enter a company as a third party to perform an audit on that data. With this, we can tell those companies which data about them is in order and which is not.”

What consequences can bias have?

“The danger of not having your AI system checked is that you are likely to wrong certain people, such as the toeslagenaffaire. If we had sat around the table at the beginning of that process, I dare to say that this whole affair could have been prevented. I think this is a big problem that many companies have not yet become aware of. At some point, they find out about this bias, and then it’s too late. Then there are sounds that an algorithm watchdog should come, but it always comes too late. A watchdog only strikes when the burglars are already inside and have taken all your valuables.”

What is the mission of de Keuringsdienst van data?

“Our mission is that data and ethics go hand in hand. Many companies are not experienced and are afraid that they are not using their data properly. I think this is part of AI maturing. You should not think that everything is going well, and that technology will take care of it. Ultimately, people are responsible, and we must design those values ​​together with our systems, but that is a challenge. For example, this ethical aspect is not included in the curriculum of the Bachelor or the Master of Artificial Intelligence. You should always ask yourself how something comes about and what the consequences of certain developments may be.”

Do you have any tips for other leaders?

“For all people who care about AI, it is extremely important not to forget the ethical aspect of data and AI. We also tackle this with de Keuringsdienst van Data. Our industry talks too much about it but does too little about it.”

How can the AI ecosystem help you?

“The AI ecosystem can help me by taking our developer training in artificial intelligence and data or requesting masterclasses for administrators. The more people that talk about the ethical aspect of technology, the better the adoption of artificial intelligence inside the organizations will take place.”


2048 1366 Pepijn van Vugt

Pepijn van Vugt

Pepijn van Vugt is editor for who specializes in data, machine learning and artificial intelligence.

My name is HAL 9000, how can I assist you?
This website uses cookies to ensure the best possible experience. By clicking accept, you agree to our use of cookies and similar technologies.
Privacy Policy