The Interdisciplinary workshop on Computational Economics, Game Theory and Social Choice, and Machine Learning will be held on 15 May at Fuglesangs Allé.

Please sign up for the workshop, no later than 9 May 2025: LINK

Program:

08:45 – 09:00    Coffee and welcome

09:00 – 09:30    Professor Ioannis Caragiannis, Department of Computer Science, Aarhus University, Fairness in Allocation Problems

09:30 – 10:00    Associate Professor Nicola Maaser, Department of Economics and Business Economics, Aarhus University, From Square Roots to Linear Rules: Designing Optimal Voting Weights

10:00 – 10:30    Associate Professor Kristoffer Arnsfelt Hansen, Department of Computer Science, Aarhus University, Computational Complexity in Game Theory and Economics 

10:30 – 10:45    Coffee break

10:45 – 11:15    Assistant Professor Stratis Skoulakis, Department of Computer Science, Aarhus University, Online Learning and Strategic Decision making

11:15 – 11:45    Associate Professor Erik Christian Montes Schütte, CoRE, Department of Economics and Business Economics, Aarhus University, Anatomizing Machine Learning Models with Shapley Values: Insights into Forecasting and Portfolio Performance

11:45                 Lunch

Abstracts of workshop talks

Ioannis Caragiannis, Professor, Department of Computer Science, Aarhus University: Fairness in Allocation Problems. 

We will present variations of the problem of allocating indivisible items to agents with valuations for the items. We will define basic fairness concepts such as proportionality and envy-freeness and discuss their basic properties. Next, we will introduce approximate versions of these concepts, such as envy-freeness up to some/any item (EF1/EFX) and maximin share fairness (MMS). We will present examples and many open problems.

Nicola Maaser, Associate Professor, CoRE, Department of Economics and Business Economics, Aarhus University: From Square Roots to Linear Rules: Designing Optimal Voting Weights

Representatives from constituencies of different sizes make decisions using a weighted voting rule, adopting the ideal point of the weighted median among them. Each representative's preferences are assumed to align with the median voter of their constituency. In the extreme cases of binary decisions and decisions over a continuous interval, different objectives—ensuring equal prior influence for individual voters and maximizing total expected utility—both lead to the same optimal weighting rules: a square root rule when voters' ideal points are independent and identically distributed, and a linear rule when preferences are sufficiently positively correlated within constituencies. The optimal selection of voting weights when there are k discrete alternatives remains an open question.

Kristoffer Arnsfelt Hansen, Associate Professor, Department of Computer Science, Aarhus University: Computational Complexity in Game Theory and Economics 

Computational complexity theory is the field of theoretical computer science that studies the phenomenon that computational problems have intrinsic requirements for the amount of time or space required for their solution. The theory of NP-completeness, one of the main intellectual exports from computer science to other disciplines, explains computational hardness arising in practical settings of many computational tasks. This theory is however not applicable to central computational problems in game theory and economics about equilibria. We give an introduction to specialized notions from computational complexity designed to address this issue and give an overview of recent results and concerns.

Stratis Skoulakis, Assistant Professor, Department of Computer Science, Aarhus University: Online Learning and Strategic Decision making

In this talk we will introduce the online learning framework and its applications in decision making in uncertain environments. We will introduce fundamental online learning algorithms and the optimality guarantees that they can offer, commonly known as no-regret guarantees. We also present applications of the online learning framework in fundamental game-theoretic settings such as auctions and traffic systems.

Erik Christian Montes Schütte, Associate Professor, CoRE, Department of Economics and Business Economics, Aarhus University: Anatomizing Machine Learning Models with Shapley Values: Insights into Forecasting and Portfolio Performance

Understanding the individual contributions of predictors to machine learning model performance remains challenging, especially given their often black-box nature. Leveraging concepts inspired by cooperative game theory, particularly Shapley values, this presentation introduces methodologies that transparently decompose both forecasting accuracy and economic performance metrics. Using empirical examples—forecasting U.S. inflation from a large predictor set and analyzing portfolio returns based on cross-sectional stock predictions—I demonstrate how Shapley-based methods provide valuable insights, bridging interpretability and practical applications in computational economics and finance.

This workshop is sponsored by CoRE