Thiemo Fetzer | Professor of Economics

Research Module Applied Microeconomics with AI

This is the course page for the Research Module course of Applied Microeconomics with AI. The module will be a broad brush covering basics in machine learning and information economics tracing the evolution focusing on applications of regression and simple machine learning techniques, towards language modelling, neural networks and the transformer based tools such as large language models.

The course being a research module will consist of three main phases:

  1. Course/lecture phase: introduction to machine learning, generative language modeling, self sovereign AI for 4-6 weeks
  2. Project phase: Students will be allocated to groups to develop a research project for 4-6 weeks with weekly team check ins to discuss progress
  3. Project Presentation and peer feedback phase

The course will be delivered mostly as a set of lectures drawing examples from research.

Lectures
Lecture material and slides are made available to the course participants as a shared folder.

  • Week 1
    • Lecture 1: 16.10.2025 16:00-18:00 (2h) Reinhard Selten Raum/RSI 0.17
    • Lecture 2: 17.10.2025 8:30-10:00 (90m) Reinhard Selten Raum/RSI 0.17
      • Introduction to model selection and shrinkage techniques an (Slides)
  • Week 2
  • Week 3
    • Interviews
  • Week 4
    • Lecture 5: 06.11.2025, 16:00 – 18:00 (2h) Reinhard Selten Raum/RSI 0.17
    • Lecture 6:
      • Recurrent Neural Networks (Slides)
  • Week 5
    • Lecture 7 & 8:
      • Transformer Architecture (Slides)

Hands on notebooks

I have prepared and shared some notebooks on how to work with self-sovereign AI, along with some slides on this github repository.

Self Sovereign AI

I am a big supporter of self-sovereign AI along with Open Source. The technology is incredibly useful and powerful but too powerful to be left for a few transnational companies or governments to control. This is why I want to highlight many of the use cases using local large language models. There are many ways of serving a large language model.