Seminar "Selected Topics from Finance"

The seminar is open to Master students. In this seminar, we will focus on the use of generative AI in investing and asset pricing.  Please see the topics listed below to get a flavour of the seminar.

To successfully pass the seminar you need to write a paper and give a presentation. Papers can be written in either German or English and should have a length of 15-20 (team of two) or 20-25 pages (team of three). For hints on how to write a paper see our guidelines. You need to hand in a printed and a digital version (pdf) of your paper, as well as data and code for your empirical analysis. The seminar talks should be given in English.

In case of further questions please feel free to contact Christof Ganzhorn.

• Do we get a grade? Yes. Your paper and your presentation will be graded and lead to one grade (equally weighted). Both the paper and presentation have to be passed.
• What do we have to hand in? A paper (around 10 days before the presentation) and your presentation files.
• Who is responsible? For content-related questions, please contact your supervisor.

• Mon. 26.01.2026 - Thu. 29.01.2026: Students must submit their preferences over seminars for the first matching round. Link to the platform for submission of preferences:  English     German
• Fri. 30.01.2026: 1st round of seminar matching. The matching algorithm runs during that day.
• Wed. 04.02.2026: 2nd round of seminar matching. The matching algorithm runs during that day.
• tba  Topic allocation on Taddle
• tba  Registration at the Higher Services Portal
• tba  Contact your supervisor to discuss the outline of the paper
• tba  Submission of the paper

• tba  (most likely in June)  Presentations 

   

The topics already listed below give you a good idea of the type of topics available in this seminar. We will add more topics before the topic allocation, which happens after seminar places have been allocated. 

General remark: The topics will include a practical task for which you will work with ChatGPT or another generative AI tool. Most tasks will also involve financial data and will require integrating outputs from generative AI into a standard financial analysis, such as portfolio construction and performance evaluation. We will provide you with hints on how to collect the data and what to do specifically as part of the practical task. You can choose the software you use to analyze the data. In your presentation, you shall provide insight into your use of gen AI and your coding approach,  Do not ask the authors of the original papers whether they can provide you with code or other assistance. We expect you to prodce the code yourself (seeking help from gen AI for that purpose is ok), and we expect all members of a team to be familiar with the way in which gen AI was used for the task, as well as any data processing and coding involved. 

 

1. ChatGPT in Systematic Investing 

Give a summary of the analysis performed in Anic et al. (2025), which uses ChatGPT to enhance momentum strategies. Then do an analysis along the lines of what is done in the paper. Given that the compilation of news data may require too much time for a seminar paper, an feasible alternative may be to ask the gen AI tool to use more detailed historical return information in addition to the standard 12 month performance. For example, existing research shows that specific months or patterns withing the 12-month window are particularly helpful for constructing momentum portfolios. It may therefore be interesting to examine whether gen AI tools can help process this or other information. 

Literature: Anic, N., Barbon, A., Seiz, R. and Zarattini, C., 2025. ChatGPT in Systematic Investing--Enhancing Risk-Adjusted Returns with LLMs. arXiv preprint arXiv:2510.26228. 

Supervisor: 

 

2. What Does ChatGPT Make of Historical Stock Returns?

 Give a summary of the analysis performed in Chen et al. (2025), which examines return forecasts produced by ChatGPT. Then do an analysis along the lines of what is done in the paper. No need to examine human forecasts as part of the practical task. 

Literature: Chen, S., Green, T. C., Gulen, H., & Zhou, D. (2025). What does ChatGPT make of historical stock returns. Extrapolation and miscalibration in LLM stock return forecasts. Available online at: ssrn. com/abstract, 4941906. 

Supervisor: 

 

3. Behavioral Biases of AI 

Decision biases play an important role in behavioral finance. Interestingly, gen AI responses exhibit some of the famous biases that have been documented for humans. Give a summary of the literature. Then do an analysis of your own, by asking ChatGPT and other gen AI tools questions that check for biases. 

Literature: Bini, P., Cong, L.W., Huang, X. and Jin, L.J., 2025. Behavioral Economics of AI: LLM Biases and Corrections. Available at SSRN 5213130. 

Supervisor: 

 

4. Lookahead bias

When ChatGPT or other gen AI tools are used for forecasting and we check in a backtest how well the forecasts would have performed, we should be aware of lookahead bias: Maybe a forecast made today by ChatGPT for returns in 2015 are good because ChatGPT knows the past realization or other information that is helpful for making the “forecast”. Give a summary of the literature: What does it say about the magnitude of lookahead bias and potential remedies? Then, using strategies from the literature, check yourself with ChatGPT and other gen AI tools whether there is lookahead bias. 

Literature He, S., Lv, L., Manela, A. and Wu, J., 2025. Chronologically Consistent Large Language Models. arXiv preprint arXiv:2502.21206. 

Supervisor:

Prof. Dr. Gunter Löffler

Christof Ganzhorn

Stefan Rausch

 

see timetable

This seminar is open for Master students.