Work

The Use of Artificial Intelligence in the Public Sector and Transparency Issues

Year: 2023
Type: dissertation
Abstract: This thesis aims to discuss the use of artificial intelligence (AI) in the public sector and related transparency issues.The types of AI which will be focused upon are machine learning and deep learning, in particular as used in the public sector fields of tax administration and radiology.The method used is the qualitative secondary research method where previous research has been analysed and drawn conclusions from, in particular, academic research and literature, legislation, case law, publications from bodies of the European Union and the United Nation, and to a limited degree reliable internet sources.The first main topic being discussed is how the use of AI in the public sector may lead to a lack of transparency.Here it is examined whether the provisions regarding the right to explanation in the General Data Protection Regulation offer adequate protection.Then two examples of so-called black box AI in the public sector are analysed, namely the use of AI to detect tax fraud by tax authorities in Poland and the use of AI in the field of medical radiology.The next main topic analysed is the consequences of opaque AI and in particular how this may prejudice the right to a fair trial, cause discrimination and a generate general lack of trust in AI tools and public decision-making.Certain positive aspects of black box AI will also be considered, such as a potentially increased efficiency, accuracy and security in certain areas where it is used and how decisions generated by AI even in certain instances can result in an increased trust by the general public despite a lack of full transparency.The conclusion is that while it is important to weigh against each other the different advantages and disadvantages of black box AI, it is not an easy task to strike a balance between the use of such AI and full transparency.The third and final main topic discussed in this thesis is whether a double standard of transparency requirements for AI tools versus humans exists, whether a higher standard imposed upon the use of AI tools may be justified and whether such a double standard actually exists.The conclusion reached is that in some cases such a double standard, or at least a demand for higher standards when AI is used, may well be justified.No such double standard however currently exists; it is rather so that humans are subjected to higher standards when it comes to explainabilty.Finally, a conclusion summarising the different findings in more detail is provided at the end of this thesis.
Author: Rosell, Catharina
Institution: Örebro University
Language: English
Subfield: Law
Open Access status: green
OpenAlex ID: W7016115866
Cites: 21
Cited by: 1
Related to: 9