How does a high population growth rate affect the level and growth of GDP? This is a common question in university economics courses, designed to train students in economic reasoning and argumentation. However, grading and giving feedback on such open-ended responses is time-consuming. Instructors must carefully assess each answer individually. Could this task be handled by an AI in the future?
This question is addressed in a recent study published in the renowned journal Scientific Reports. The research team, consisting of economists and computer scientists from the University of Passau, faced a key challenge: not only can AI make mistakes, but so can human graders. If an AI’s assessment differs from that of a human, it’s not immediately clear who is at fault. So how can one make a comparison when there is no definitive ground truth?
The study uses a clever approach: the level of agreement among multiple human graders is taken as an indicator of accuracy. Using a dataset of student responses from actual university practice, the researchers replaced one of several human graders with an AI. If the overall agreement dropped, this would indicate lower AI quality.
Surprisingly, the results showed that GPT-4, as an AI model, could at times match the performance of human graders. This was especially true when the task involved ranking responses—i.e., identifying the best, second-best, and worst answers—rather than assigning absolute point scores. GPT-4 also showed no bias toward longer responses or AI-generated content.
So, can human graders be replaced in the future? Not quite. Model answers and final reviews must still be handled by humans. Another caution emerged from the study: in tasks requiring point-based scoring instead of ranking, GPT-4 performed somewhat worse. The authors therefore argue that exams should continue to be closely supervised by humans. However, GPT-4 can be a valuable second assessor, and wherever feedback is needed without assigning final grades, its performance is impressively fast and helpful.
The article is available at the following link: 10.1038/s41598-025-21572-8
In a recent publication in Ecological Economics, Stephan Geschwind and Johann Graf Lambsdorff explore how resource scarcity influences human behavior. They conduct a novel lab experiment that simulates an environment with exploitable resources to study the extent of hostility. The study finds that hostility increased towards players responsible for overexploiting resources, i.e., when scarcity was human-induced. Conversely, environmentally induced scarcity mitigated hostility. These findings offer important insights for policymakers. They suggest that the climate crisis may escalate or mitigate hostility and violence, depending on whether it is perceived as being caused by human actions or natural factors.
The article can be accessed via the following link: https://doi.org/10.1016/j.ecolecon.2024.108388
The corporation became the dominant form of firm in the 19th century, but its nature and advantages are still poorly understood. A recent article "The advantages of the corporate form - an impossibility theorem on persons and things" by Johann Graf Lambsdorff in the Cambridge Journal of Economics provides a game-theoretic proof of these advantages. To achieve these, the corporation had to be given legal capacity, treating it as a person, and be transferable, treating it as a thing. A classic dichotomy between persons and things. This dichotomy had to be overcome in order to secure the advantages of the corporation.
The article can be accessed via the following link: https://doi.org/10.1093/cje/beae003
Do people lack willpower when they don't stick to their long-term resolutions? In collaboration with researchers from the United Kingdom, Dr. Kevin Grubiak finds in a new study that subjects enjoy the freedom to spontaneously deviate from their resolutions. The results have implications for the assessment of interventions in the spirit of libertarian paternalism, which have so far neglected the intrinsic value of spontaneity.
The study was published in the journal Behavioural Public Policy and can be accessed via the following link: https://doi.org/10.1017/bpp.2021.41
The Internet Center for Corruption Research explores corruption worldwide, with a particular focus on behavioral perspectives. In addition to data, indices, and studies, it offers courses and workshops for students and professionals.
The chair does research in the areas of Experimental and Behavioral Economics, Economics of Corruption and Experimental Macroeconomics.
Publications of Prof. Dr. Johann Graf Lambsdorff in PDF and on Research Gate
Publications of Dr. Kevin Grubiak
Publications of Dr. Katharina Werner and on Research Gate
Publications of Prof. Dr. Marcus Giamattei on Research Gate
