Programme of the summer school

The Summer School consists of five tutorials, addressing synergies and mutual benefits between Data Science & AI and Economics. They are delivered by leading experts in their field.

Each tutorial will cover half a day, including a theoretical introduction and followed by a practical part.

The tutorials are aimed at PhD students, postdoctoral researchers, and early-career researchers with a basic knowledge in Data Science or AI. Motivated MSc students are also welcome!

Knowledge of Economics is advantageous but not required as each speaker will briefly review the necessary topics.

A general interest in interdisciplinary applications of Data Science is highly recommended. The summer school encourages active networking between participants and speakers.

Schedule (last update 14.05.2026)

Speakers

The tutorials will be delivered by the following speakers. You will find their abstracts in the next section.

Andreas Haupt, Stanford University (USA)

Tutorial: Economics in Machine Learning from Human Preferences

Despoina Makariou, University of St.Gallen (Switzerland)

Tutorial: Machine Learning and Data Science for Risk and Insurance

Massimiliano Marcellino, Bocconi University (Italy)

Tutorial: Machine Learning Based Macro-Econometrics

Larysa Zomchak, National University of Lviv (Ukraine)

Tutorial: Complexity, Chaos, and Machine Learning in Economics: Tools for Understanding Unpredictable Systems

Behnud Mir Djawadi, Paderborn University (Germany)

Tutorial: Human-AI Interaction in Economic Experiments – How AI Systems Shape Human Behaviour

Abstracts

Abstracts of all tutorials for the Summer School 2026 are listed below.

* Andreas Haupt: Economics in Machine Learning from Human Preferences

The tutorial will highlight ideas from the behavioral sciences in Large Language Model training. Pre-training uses proper scoring rules to elicit calibration, which disappears in mid- and post-training due to reinforcement learning. Supervised finetuning matches moments. Reinforcement learning from human feedback uses discrete choice methods, and aggregates preferences according to the Borda count. Participants will leave with an overview of current training methodology, and a list of open problems for future work for Economists and Data Scientists in the training of human-centered language models.

* Despoina Makariou: Machine Learning and Data Science for Risk and Insurance

This tutorial introduces how machine-learning and data-science methods can be applied to central problems in risk and insurance, which play an important role in economic and financial decision-making. The theoretical part will review core modelling concepts such as claims frequency and severity modelling, predictive approaches, and model evaluation, highlighting how these tools extend classical actuarial and econometric techniques. The tutorial will also cover key practical considerations, including data preparation, interpretability, and the challenges typical of insurance datasets. The hands-on component will demonstrate simple, reproducible workflows using public or simulated data to illustrate how machine-learning methods can be applied in insurance and risk-management contexts.

* Massimiliano Marcellino: Machine Learning Based Macro-Econometrics

A variety of Machine Learning methods have been recently imported into Macro-econometrics to allow for the use of both very large data sets and more flexible functional forms. We will review some of the resulting classical and Bayesian Machine Learning based macro-econometric methods, focusing on those that have been shown to produce good forecasts for economic and financial variables, with respect to standard macro-econometric models. We will use various empirical applications to illustrate the methods.

* Larysa Zomchak: Complexity, Chaos, and Machine Learning in Economics: Tools for Understanding Unpredictable Systems

Economic systems are complex, nonlinear, and often chaotic, yet mainstream econometric practice has long relied on linear models that struggle to capture this reality. This tutorial revisits the foundations of econometric modelling through the lens of chaos theory, and explores how modern Data Science methods can extend, rather than replace, the econometric tradition.

The first part of the tutorial introduces chaos theory as an economic modelling framework: deterministic chaos, nonlinear dynamics, sensitivity to initial conditions, and strange attractors. Drawing on examples from business cycles, financial markets, and macroeconomic time series, participants will develop an intuition for why economic data so frequently defies classical model assumptions, and what chaos theory offers as an alternative perspective.

The second part focuses on non-linear econometric methods: threshold models, regime-switching models, GARCH-family volatility models, and cointegration in nonlinear settings. These tools are examined both theoretically and through applied examples, including GDP forecasting, inflation dynamics, and banking sector modelling. Special attention is given to the practical challenges of working with real economic data — structural breaks, mixed-frequency observations, and limited sample sizes.

The third part reflects on the role of Machine Learning (ML) as a complement to econometrics in chaotic economic environments: where ML adds value, where it falls short, and how hybrid approaches combining econometric rigour with ML flexibility are emerging at the research frontier.

This tutorial will blend theory, applied examples from European and transition economies, and open discussion.

* Behnud Mir Djawadi: Human-AI Interaction in Economic Experiments – How AI Systems Shape Human Behaviour

The tutorial explores how humans interact with AI systems when making economic decisions under controlled experimental conditions. Drawing on recent behavioural research, we examine how exposure to algorithmic advice, chatbots, and other AI agents influences human preferences, trust, cooperation, and strategic behaviour. Participants will be introduced to the method of experimental economics used to isolate these effects and to the key behavioural insights emerging from this growing literature. The session also includes a hands-on component in the form of several live in-class experiments, giving participants first-hand experience of human-AI interaction as both research subject and object of study. In one experiment involving a creativity task with a chatbot, participants can even earn some prize money.