What are recommender systems?

Recommender systems shape our digital interactions more than we realize. Every time Netflix suggests your next favorite show, or Spotify curates a playlist that seems to read your mind, a recommender system is at work. But what exactly are these systems, and why are they so important?

In simple terms, a recommender system is an AI-driven tool that predicts user preferences based on historical data and behavioral patterns. But making accurate recommendations is far from easy. These systems must balance personalization, diversity, fairness, and user satisfaction while considering the ethical implications of filtering information for people.

Martijn, an expert in the field, points out an interesting challenge:

"A common issue with recommender systems is that they rely too much on historical data. But what if we could help people break out of their existing preferences and explore something new?"

This leads us to his personal journey into data science.

How did Martijn get into data science?

Every data scientist has a unique path into the field, and Martijn’s journey is no exception. His story begins at Eindhoven University of Technology (TU/e), where he studied Electrical Engineering. While his initial focus was on traditional engineering, he soon discovered an interest in how technology affects decision-making. This curiosity led him to Technology & Society, where he started conducting empirical studies on cognitive processes and decision-making.

His fascination with how people process information and make choices led him to research on process-tracing methodologies. His work expanded into tracking user decision making with his own developed online process tracing tool (MouselabWEB), using statistical models to analyze decision-making.

Martijn recalls,

“I just liked doing experiments. I wanted to understand how systems work for USERS, not just from a technical perspective, but from a human perspective."

This shift in thinking naturally led him into recommender systems and user experience evaluation. Over the years, he became a recognized expert in the field, even organizing the RecSys Conference in Amsterdam in 2021, where industry professionals and researchers shared insights on the future of recommendation technology.

However, with the rise of deep learning and AI, Martijn has some concerns. He warns that many people are overlooking the importance of real statistics and user-focused evaluation.

“Optimizing a single metric doesn’t tell you if users are truly satisfied. We need a deeper understanding of human behavior beyond just algorithms.”

This realization became even more evident during one of his most exciting projects—an experiment that combined data science and classical music.

The JADS music festival project: a data-driven opera experience

Imagine walking into a concert where the music is tailored to your Spotify profile—a personalized opera experience, blending technology with classical music. Sounds futuristic?

Well, this project became a reality at JADS (Jheronimus Academy of Data Science), where Yu and Martijn collaborated on her PhD project to develop a system that matched classical music performances to people's musical tastes. The idea originated when the International Vocal Competition 's-Hertogenbosch reached out to them, seeking a way to blend data science with classical music. Inspired by the challenge, Yu and Martijn designed an algorithm that would not only recommend music based on historical preferences but also encourage listeners to discover new genres through a personalized concert experience.

Here’s how it worked:

- Participants were asked to link their Spotify profiles before attending the event.

-The system analyzed their listening habits, favorite genres, and mood-based preferences.

-Using this data, an algorithm matched them with opera singers’ repertoires to create a unique listening experience.

-Instead of reinforcing their usual tastes, the system encouraged them to explore new but personalized genres.

-Attendees provided real-time feedback, helping refine future recommendation models.

The goal? To prove that recommender systems can not just reinforce what you already know—they can help you explore new interests.

Martijn explains: "People don’t always know what they like. They discover their preferences while making decisions. That’s why recommender systems should guide users, not just reflect back their past choices."

This project was an incredible success, and Martijn is already thinking of expanding it to different music styles. What if AI could help you fall in love with a genre you’ve never considered before?

Current trends in Data Science – are we losing analytical skills?

What does the future of data science look like? Martijn has a mix of optimism and concern. He believes that while deep learning has revolutionized AI, we are losing something crucial—critical thinking and deep analytical skills.

"With all the hype around LLMs (large language models), people say ‘oh, they can do this and that, they can even reason!’ But I don’t actually think so. These models don’t really understand the world—they just predict patterns."

His concerns go beyond just AI hype:

• APIs from companies like Spotify have enabled powerful research, but there is growing fear that companies will start restricting data access.

• Are we too focused on automation and not enough on real understanding?

• Are companies adopting AI without questioning its real-world implications?

"I’m worried that we are no longer learning. If we blindly rely on AI without understanding the foundations, we will lose essential competencies."

Even in recommender systems, a field deeply tied to AI, there is growing evidence that simpler models often outperform deep learning approaches. Researchers are realizing that sometimes basic methods are more efficient and effective than complex, resource-intensive deep learning models.

The real takeaway? Critical thinking should always come first in data science.

Final thoughts

Martijn’s journey—from decision-making research to cutting-edge recommender systems—highlights the importance of understanding human behavior alongside AI advancements.

While the future of data science is exciting, his insights remind us that technology should serve users, not the other way around. Whether it’s helping people discover new music, designing ethical AI, or questioning industry trends, Martijn believes that data scientists should always think beyond the algorithm.

His message is clear: stay curious, question everything, and use AI as a tool, not a replacement for deep understanding.

And that’s what makes this field so fascinating.