Choosing the Agent: An Interview Study on Reinforcement Learning Projects in Industry

Forschungsthema/Bereich

Reinforcement Learning · Interview Study

Typ der Abschlussarbeit

Master

Startzeitpunkt

-

Bewerbungsschluss

31.08.2026

Dauer der Arbeit

6 Monate, ab sofort

Beschreibung

Choosing the Agent: An Interview Study on Reinforcement Learning Projects in Industry

Reinforcement learning is one of the harder machine learning methods to get working outside a simulator. Trial and error is fine in a lab, expensive in a factory, and rarely something a product owner will sign off on. And yet companies do build it.

RL is prominent in research and hard to spot in the wild, partly because it is almost never a product in itself but sits somewhere inside one. Some initiatives presumably ship and run quietly for years. Others may stop at the proof of concept, or run for a while and then get replaced by a simpler method that is almost as good and far easier to operate. Which of these is the common case is not documented anywhere, and neither are the reasons.

The technical challenges of real-world RL are well described in the literature. The organisational side is not. And it matters beyond RL: this is the case where the familiar picture of machine learning adoption, a model predicts and a person decides, no longer holds. Here the system decides by itself, over and over, and its behaviour shifts as the policy changes.

In this thesis you talk to the people who were there. You reconstruct real RL initiatives without assuming how they ended: why RL was chosen over something simpler in the first place, how development actually ran, which challenges and risks came up along the way, what the alternative would have been, and what became of the project.

Specifically:

• Identifying and recruiting practitioners yourself, e.g. through patent inventors, companies, startups. We will support you here.
• Reconstructing initiatives across whatever range of outcomes you find, from systems running in production to projects that stopped early or were later replaced
• Designing and running semi-structured expert interviews
• Qualitative coding with a deductive starting scheme from the real-world RL literature, extended inductively from the material
• Developing a taxonomy of industrial RL initiatives, treating what became of a project as the thing to be explained rather than as one more attribute

This is not an implementation thesis. You will not build, train or benchmark an RL system. The contribution sits at the organisational and socio-technical level.

What you bring:

• an interest in qualitative empirical research
• the persistence to approach practitioners and run the interviews yourself
• enough machine learning background to hold a technical conversation and to tell RL apart from neighbouring methods. You do not need to be an RL expert
• confident German and English

What you get:

• support with identifying and approaching interview partners
• close supervision on study design, interview guide and coding
• direct contact with the people building RL systems in industry
• a genuine empirical contribution that feeds into ongoing research at the chair

Disclaimer
The thesis is meant to teach you the scientific approach. Please keep that in mind. We expect you to learn and use scientific methods and to submit a scientifically sound result. Our role is to provide you with everything you need to succeed at that. We expect that you create a piece of work, in which you demonstrate a sound data collection method, an extensive dataset and your analysis of that dataset. A motivated student who works consistently on the thesis throughout the designated period can expect a very good grade. Details on the process and the format that we use to support you in succeeding with your thesis can be found here:

Master: https://bwsyncandshare.kit.edu/s/MiN8yTkfgeCrnSW

If you are interested, please send a cover letter and transcript of records by email to julius.grams@kit.edu.

References

• Dulac-Arnold, G., Levine, N., Mankowitz, D. J., Li, J., Paduraru, C., Gowal, S., & Hester, T. (2021). Challenges of real-world reinforcement learning: definitions, benchmarks and analyses. Machine Learning, 110, 2419-2468.
• Myers, M. D., & Newman, M. (2007). The qualitative interview in IS research: Examining the craft. Information and Organization, 17(1), 2-26.
• Nickerson, R. C., Varshney, U., & Muntermann, J. (2013). A method for taxonomy development and its application in information systems. European Journal of Information Systems, 22(3), 336-359.

Voraussetzung

Voraussetzungen an Studierende

• An interest in qualitative empirical research
• The persistence to approach practitioners and run the interviews yourself
• Enough machine learning background to hold a technical conversation and to tell RL apart from neighbouring methods. You do not need to be an RL expert.
• Confident German and English

Studiengangsbereiche

• Wirtschafts- und Rechtswissenschaften
  Wirtschaftsinformatik

Betreuung

Titel, Vorname, Name

Julius Vincent Grams

Organisationseinheit

KIT WIN - ESIS

E-Mail Adresse

julius.grams@kit.edu

Link zur eigenen Homepage/Personenseite

Website

Bewerbung per E-Mail

Bewerbungsunterlagen

• Anschreiben
• Notenauszug

E-Mail Adresse für die Bewerbung
Senden Sie die oben genannten Bewerbungsunterlagen bitte per Mail an julius.grams@kit.edu


Zurück