Management

AI identifies corruption risks in the public sector

30/06/2026

New study shows: Public servants’ values and attitudes predict corruption susceptibility better than income or education.

WU researchers are leveraging AI to analyze susceptibility to corruption among public servants worldwide. The results show: Individual attitudes towards democracy, competition, and leadership are more relevant than education or income.

Values and attitudes are key

Research on the causes of corruption has so far largely focused on institutional, demographic, and cultural determinants as well as differences between countries. However, the role of individual attitudes and beliefs in shaping public servants’ susceptibility to corruption has received little systematic attention. WU researchers Moritz Schmid and Jurgen Willems have now used artificial intelligence to examine the extent to which corruption susceptibility can be predicted based on individual attitudes and which factors play a key role. “Our results challenge a common assumption: It’s not income or education that best predict corruption susceptibility among public servants, it’s their values and attitudes,” says Moritz Schmid from WU Vienna’s Institute for Public Management & Governance. The findings clearly show that a strong alignment with democratic values tends to be associated with lower corruption susceptibility. “Individuals with a strong commitment to democratic values show less tolerance towards corrupt behavior,” Schmid explains.

A global dataset and multiple AI models

Portraitfoto von Moritz Schmid

Dr. Moritz Schmid, Institute for Public Management and Governance

© Lars Ternes

The findings are based on data from 18,000 public servants across 90 countries, covering their cultural values and attitudes on topics such as family, religion, social tolerance, and trust in institutions. The data do not capture actual corrupt behavior. Instead, corruption susceptibility is assessed based on attitudes, by asking the respondents whether bribery, tax evasion, and the improper claiming of public benefits are justifiable. “It’s difficult to observe corrupt behavior at the level of the individual. That’s why we rely on attitudes to measure corruption susceptibility,” says Schmid.

New insights through explainable AI

Potraitfoto von Jurgen Willems

Prof. Jurgen Willems, Institute for Public Management and Governance

© Jurgen Willems

In addition to classic regression models, the research team used several explainable machine learning approaches to predict public servants’ corruption susceptibility. The models not only achieved higher predictive accuracy but also enabled a systematic comparison of the importance of more than 100 potential influencing factors. “Machine learning can reveal relationships that often remain hidden when using traditional approaches,” says Schmid. “Beyond its substantive findings, the study also provides new methodological avenues for researching integrity, and it highlights potential anti-corruption measures in the public sector,” Willems adds.

Detailed study results

Schmid, Moritz, Willems, Jurgen (2026): AI for Integrity: Predicting Public Servants' Susceptibility to Corruption via Supervised Machine Learning. In: Government Information Quarterly (2026). Available at: https://doi.org/10.1016/j.giq.2026.102158

Back to overview