The aim of the project was to develop a trustworthy AI component that supports the whistleblowers and the clerks of a whistleblowing system through partial automation.
Employees or other persons involved often do not report violations such as non-compliance with internal guidelines, bullying or sexual harassment for fear of negative consequences. These can include discrimination, sanctions or loss of job. A trustworthy reporting channel (whistleblowing system) is therefore essential. The focus here is on receiving high-quality reports that can be used to quickly clarify an incident, as well as protecting the whistleblower. However, the evaluation of reports via conventional reporting channels often means a very high manual effort for the responsible caseworkers, as the quality of the report texts is rarely sufficient and at the same time the whistleblowers only have a limited amount of time available for queries. Companies are reluctant to implement a whistleblowing system due to the high costs associated with the resources required to analyze the reports or the misuse of the information to spread false reports or to spread corruption.
EU Directive 2019/1937 on the protection of whistleblowers is to be transposed into national law by December 2021. Compliance 2b GmbH is a start-up founded in 2020 to address this issue. The Act4Whistleblowing web solution developed enables a secure, anonymous exchange of information between whistleblowers and companies as well as simple and innovative processing of reports by employees. In a further expansion stage of the platform solution, RISC Software GmbH, together with Compliance 2b GmbH, is developing a trustworthy AI component as part of a research project, which on the one hand supports the whistleblowers in submitting the report through partial automation and on the other hand should enable the responsible clerks to process the reports more efficiently and less error-prone.
The system automatically derives information from the textual information and compares it with data that has already been structurally recorded in the form of a plausibility check. This automated check opens up completely new possibilities in this process, such as requesting further essential information from the whistleblowers in real time and supporting the case handlers by pointing out potentially incorrect information or results from similar cases. The increased quality of reports increases the chances of successfully resolving cases and reduces the likelihood of reprimands.
The research project was funded by Austria Wirtschaftsservice (“aws”) as one of 12 out of 68 submissions in the second round of the call “Digitalization – Trustworthy Artificial Intelligence”
Data Scientist
