by Aslı Aydoğan
During my stay at the Netherlands Institute in Turkey (NIT), I worked on the early stages of my Master’s thesis. My research looks at what good governance should look like when governments use algorithmic risk models in social assistance, with lessons from the Dutch RCM case (2014–2020) and a focus on Turkey.
Why Turkey, and why social assistance?
Algorithms increasingly influence public decisions worldwide, especially in social assistance, where outcomes have a direct impact on people’s lives. While Western Europe has been studied extensively, much less is known about how countries like Turkey are digitalizing their public sectors. With the Ulusal Yapay Zeka Stratejisi (2021–2025) and the 2024 report Kamuda Yapay Zekâ Uygulamaları, Turkey is actively exploring AI in public services. This creates an important moment to ask: What should be in place before algorithmic risk tools are rolled out?
What I worked on at NIT
NIT offered the perfect environment to explore both Dutch and Turkish policy contexts. I focused on three things:
- Re-examining the Dutch RCM case: how it worked, why it failed, and what governance issues played a role.
- Mapping Turkey’s current AI and digitalization policies.
- Designing my interview guide and identifying the experts I hope to speak with.
What I’ve learned so far
- Governance matters more than the technology itself. Algorithms reflect the culture, data practices and legal frameworks they operate in.
- Turkey is at a turning point. Unlike in the Netherlands, where systems were implemented before oversight was fully in place, Turkey can still design safeguards before moving forward.
- Many questions remain. How do you protect citizens’ rights when decisions become automated? What types of information are necessary to properly evaluate such systems? How can different actors within the public administration shape their roles and responsibilities around AI applications?
These questions will guide the next steps of my research.
What’s next?
In the coming period, I will focus on deepening my understanding of the Turkish context and gathering expert insights. Based on this, I will analyze which governance conditions are necessary to ensure that future algorithmic risk classification in social assistance can be implemented responsibly.
My time at the NIT has provided a strong foundation for this research. It was an inspiring place to begin this project, and I am grateful for the support and space I received. In the months ahead, I will continue developing my research and further elaborating on my findings.