Moving from Machine Learning (ML) to the Intersection of Economics and ML

27 Dec 2025 • Personal

In the last few months, at least 8 people have asked me in surprise why I moved from pure machine learning into economics. Their reasoning was along the lines of why when AI is booming I am sidestepping and pivoting to something else.

Hence, I thought I would write down my thinking, partly to have something to point people to, partly to put down my thoughts for my future self.

For context: I recently started an interdisciplinary PhD at ETH Zürich. I have always had a keen interest in economics; if I were able to double major, econ would have been that second major. Hence, it was already clear for me wanting to do something interdisciplinary between these two subjects. However, besides my interest and fascination, I took quite a lot of time in December of 2024 to reflect - originally I thought about enrolling in a pure machine learning PhD but feeling fundamental changes coming to the traditional ML research career path made me focused on an interdisciplinary PhD.

If at the end of 2024 I had been asked what my core talent was, I would have said coding. I was by no means the best coder but I was able to push out decent code at a very rapid pace - I am a builder at heart.

However, the days of endless computer science growth are over.

The U.S. Bureau of Labor Statistics projects computer programmer employment to decline 10% from 2023 to 2033, explicitly citing AI automation as the reason.

On the research side, NeurIPS submissions went from around 1,000 in 2014 to over 13,000 in 2024. That’s 13x more competition for attention, and you’re competing with billion-dollar labs.

As a field gets more crowded, an individual’s marginal contribution shrinks.

That said, economics has its struggles too: The econ PhD job market is not faring too well. Job postings are down 30% in three years. Tenure-track openings dropped from 631 in 2022 to roughly 400 in 2025. Last year, 5341 candidates competed for these positions with only 99 fresh PhDs secured tenure-track jobs in America, equating to a 7% placement rate.

US data suggests the traditional exits are closing too: federal hiring freezes, IMF and World Bank freezes, tech companies getting leaner.

So why focus on machine learning and economics? The intersection of ML and economics is sparsely populated. The pipeline flows almost entirely the other direction: economists learning machine learning tools, not computer scientists learning economics.

For me concretely, I would want to leverage this opportunity to bring microeconomic methods into machine learning. Many concepts in modern AI could benefit from an economic lens: the evaluation of LLMs is increasingly becoming a social science, alignment of models is essentially trying to specify an aligned utility function onto LLMs, and mechanism design offers frameworks for multi-agent AI systems.

Hence I am positive that post-hoc this will have been a good idea, I hope I can revisit this post in 5 or 15 years again and reflect on it. :)

Have any feedback?

Please feel free to send me a mail! I would love to hear from you.