This contribution looks at the ways in which machine learning technologies linked to public sector finance are creating new forms of politics and corruption thus reshaping/reconfiguring relationships across online and offline arenas. This is based on the case study of Robodebt which refers to a controversial automated debt recovery program implemented by the Australian government, particularly by the Department of Human Services (now Services Australia), which began in 2016. The program was aimed at identifying and recovering over-payments made to welfare recipients, primarily through the country's social security system. From 2016 to 2019, the Robodebt scheme raised more than half a million inaccurate debts through a method of 'income averaging', which has since been ruled unlawful and nearly $2 billion was given in settlements by Australian courts for victims of government’s failure. The article uncovers the ways in which machine learning is disrupting social and economic relationships in both public and private sectors. It reveals a world of unanticipated consequences of algorithmisation that spans online and offline arenas.

Digital governance is gaining increasing interest in public administration because it enables public value creation in a more responsive and accountable manner (Larsson and Skjølsvik 2023; Valle-Cruz 2019; Twizeyimana and Andersson 2019; Meijer et al. 2018). It holds promise for public organizations to improve service quality, reduce bureaucracy, increase citizen participation, formalize the rules and increase equality, fairness, and transparency (Larsson 2021; Twizeyimana and Andersson 2019; Bannister and Connolly 2014). For this reason, researchers have chosen to investigate the impact of digital governance on reducing the risk of corruption.

In the case of corruption, research indicates that the level of corruption detection in the public sector increases due to greater financial transparency (Puron-Cid et al. 2019), some of which is pursued through digitalization initiatives. Indeed, technological development is the best predictor of corruption control, according to Androniceanu et al. (2022). However, Mutungi et al. (2019) argue that corruption depends primarily on conditions conducive to it and social predispositions, which cannot always be controlled and moderated through digital technologies. In the same vein, Valle-Cruz (2019) add that "it is complicated to confirm that corruption is reduced using technologies" (p. 542), although there is a chance for this, e.g. using AI. At the same time, some evidence from the rapid digitalization which occurred during the COVID-19 pandemic suggests that digitalization can open newer doors to corruption (Polzer and Goncharenko 2022). Finally, the results of a study by Mouna et al. (2020) suggest that there is no direct relationship between digitalization and corruption but, rather, mediated by economic growth.

Due to the lack of consistency in researchers' opinions, as mentioned above, showing digital governance pitfalls and regulatory hurdles in public organizations, this paper aims to determine how digital governance affects corruption and ethical decision-making in public organizations. To achieve this aim, we used Flyverbom's (2022) digital prism as a conceptual framing.

We provide evidence based on the results of the preliminary exploratory study derived from 21 individual in-depth Interviews conducted with purposefully sampled public servants in Poland from local government, education, police, anticorruption bureau, and fire brigade. In doing so, we are responding to the challenge thrown by Agostino, Saliterer and Steccolini (2021) to advance knowledge of the effect of digitalization on accountability and transparency, particularly in less-studied countries.

We conducted interviews in the second half of 2023. The results present the public servants' and experts' opinions about the impact of implementing information technologies (IT) on corruption in Poland. We analyzed the possibilities created by IT, identified public administration problems that can be reduced thanks to IT, examined the attitude of public servants towards IT, and analyzed the relationships between IT use and public service transparency. Our findings thus far enrich the digital governance theory, pointing out barriers and opportunities for digitalization to promote good governance and reduce corruption.