Long Hoang Nguyen

Motivation:
Artificial intelligence (AI) is becoming increasingly important for organizations and society due to new opportunities in terms of automation and innovation. As AI systems become more widespread, scandals highlight various problems with contemporary AI systems, such as biases leading to discrimination. Overall, users of AI systems currently face a variety of problems related to ethics, data protection, and cybersecurity. To overcome these problems, it is becoming increasingly important to know who can be held accountable for the adverse consequences of AI use. In other words, AI accountability must be established to ensure that perpetrators are punished and those affected are compensated. However, there are currently numerous challenges for AI accountability. Among other things, the mechanisms behind AI systems are opaque. The overall complexity of AI systems also makes it difficult to achieve AI accountability. As the development of AI systems progresses (recent example: OpenAI Sora), new challenges are constantly being added. Gaining an overview of the current challenges of AI accountability would be a first step toward a potential solution.

Objectives:

The aim of the work is to identify and aggregate current challenges of Accountable AI.

Note: This is an umbrella topic. The overall goal, context, and direction of the thesis are defined in the first kickoff meeting.

Research Method:

Systematic Literature Review or Expert Interviews

Literature:

Bovens, M. (2007). Analysing and Assessing Accountability: A Conceptual Framework. European Law Journal, 13(4), 447-468.

Busuioc, M. (2021). Accountable Artificial Intelligence: Holding Algorithms to Account. Public Administration Review, 81(5), 825-836.

Wieringa, M. (2020). What To Account for When Accounting for Algorithms: A Systematic Literature Review on Algorithmic Accountability. In Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency (pp. 1-18).