Humboldt-Universität zu Berlin - School of Business and Economics

Writing period: 60 calendar days.

Approximate length: around 60,000 characters without spaces (about 25 to 30 pages without appendix).

A Bachelor’s thesis at the Chair of AI in Business and Economics typically follows the structure of a short research paper. While details may vary depending on the topic, a standard outline includes:

• Introduction – present the research question, motivation, and objectives
• Background and Related Work – summarize the theoretical background and briefly review existing methods or literature
• Methodology – describe the chosen method or algorithm, and any adaptations or implementation details
• Data and Empirical Study – introduce the dataset and explain how it is used to illustrate or test the method
• Results and Discussion – present findings, interpret them, and discuss limitations
• Conclusion – summarize key insights and suggest directions for future work
• Appendices (e.g., code snippets, additional figures, or tables) and references should be included as needed.

Please make sure to carefully read the faculty's recommendations related to academic writing.

For a Bachelor’s thesis, the theoretical component does not need to be substantial. Topics involving the application of AI, ML, or statistical methods to empirical data are suitable, provided the data or parameter studied is interesting and well motivated.

Writing period: 90 calendar days.

Approximate length: around 110,000 characters without spaces (about 60 pages of text without appendices)

The structure of a Master’s thesis is similar to that of a Bachelor’s thesis. However, the Methodology and/or Data and Empirical Study components must include a clear element of novelty — for example:

• new theoretical developments that extend or refine existing approaches in the literature
• innovative or comprehensive sensitivity analyses
• robust and well-designed simulation studies

A strong Master’s thesis should demonstrate originality and independent thinking. Ideally, a high-quality Master’s thesis can be developed further into a scientific paper suitable for publication in a statistics, machine learning, or AI journal.

Our chair focuses on developing new methods and algorithms in machine learning and artificial intelligence.
If your PhD interests lie in applying existing methods to economic or financial topics, please contact our colleagues who specialize in those areas.
If, however, you are interested in a PhD centered on developing new methods and algorithms, please get in touch with Katarzyna. This applies both to applicants who have already secured funding and to those who have not yet done so.

Find general information of PhDs at our school here:

FAQ of doctoral degree regulations 

Berlin School of Economics PhD program

You should be familiar with core concepts in probability and statistical inference. Prior coursework in computational statistics, econometrics, or machine learning is helpful but not required.

Being open-minded and research oriented. Some level of programming (in any language) would be beneficial, but it is not mandatory.

No, we currently accept students throughout the year.

You may want to consider that opportunities for personal meetings are somewhat limited during summer breaks due to conference attendance, research visits, etc. Therefore, if you would like close collaboration with your supervisor, consider scheduling your thesis so that the majority of the writing period falls within teaching periods. Also, before starting to write, you might want to watch this YouTube video.

It is good to plan ahead. However, there is not much that needs to be done far in advance. Please note that we allocate thesis topics on a first-come, first-served basis. This means that we do not reserve topics for students planning to start their dissertations several months later. You may, however, suggest your own thesis topic. In such cases, we can agree on a topic before the official start of the thesis.

We cannot make any recommendations related to programs offered at other universities. For Master programs at our school, you may find information on the admission requirements here.

First of all, to be eligible to register for the final thesis you have to fulfill certain requirements (e.g. you must have completed certain modules or have earned a certain amount of ECTS – depending on your degree program). To verify that, you can consult the examination office. Afterwards, obtain a thesis application form from the examination office. Fill it out and submit it to your supervisor for signature. Once signed, submit the completed registration form to the examination office – either in person during office hours or by post. You will receive a copy of the processed form once the examination office has completed the procedure.

We are primarily interested in developing new statistical, ML, and AI methods to estimate target parameters in settings with data scarcity, missingness, or structured data (e.g., causal inference, survey statistics, semi-supervised learning, missing data and imputation, hierarchical or multi-level data).

The specific method, learner, or algorithm will be defined when starting the thesis. We are very open in this respect, but we do not develop Bayesian methodology (see General Comments above).

A good starting point is to look at the topics on the Moodle page for the Research Seminar in Data Science: Learning from Incomplete Data — Causal Inference, Semi-Supervised Learning, and Beyond. If you feel inspired, please send an email to Katarzyna to discuss what you might be interested in.

Yes, you can — in fact, it is highly recommended, as working on a topic that truly interests you is more enjoyable and motivating.

The main focus of the Chair of AI in Business and Economics (CAI) is on developing new AI, ML, and statistical methods or algorithms that are versatile enough to address open questions across various domains. In other words, our research is method- and algorithm-driven, rather than driven by a specific empirical question.

Therefore, while applying existing methods to a new dataset may be suitable for a Bachelor’s thesis, it will usually not be sufficient for a Master’s thesis (unless the work includes meaningful methodological extensions) and certainly not enough for a PhD thesis with Katarzyna.

There are several possibilities to acquire the data for a dissertation. Examples include:

• The master’s thesis is written in cooperation with an industry partner who provides data
• The data is collected during (as part of) the thesis (e.g., through accessing a data provider’s API, such as Twitter, or web scraping)
• The data comes from an academic data mining/forecasting competition (KDD Cup, Data Mining Cup, NN3 or NN5 Competition, etc.) or a Kaggle competition (www.kaggle.com)

An extended abstract is typically two to four pages long and should clarify:

• Which research question(s) you plan to analyse
• The academic and practical importance of your topic
• How your thesis will contribute to the existing literature

In addition, an extended abstract should include a short list of relevant literature.

While there is no official template, we strongly advise you to look at templates of research papers published in top journals in statistics or at leading ML/AI conferences, such as Annals of Statistics, Biometrika, JRSS B, JASA, NeurIPS, ICLR, or ICML.

We currently do not have any ongoing industry collaborations. However, if you are working in a company or can facilitate a collaboration relevant to your research, this is of course possible.

The thesis should be written in English.