Mind Attribution in Regulations

Description: In previous work, we have shown that XAI researchers use mind attributing language to refer to AI systems and that this affects how participants of a survey rate awareness and responsibility of the system. It remains to be studied whether mind-attributing sentences found in XAI research "trickle down" to regulations about the ethical and legal norms for AI systems. systems.

The overall goal of this project is to determine whether the mind-attributing language usage in XAI research patterns is also found in regulations. The project consists of creating a data set of regulations (similar to [1]), applying the methods of semantic shift detection used in our previous work, analysing the results quantitatively and qualitatively. The outcome of this project is an assessment of the mind-attributing language and an investigation whether the language used in research has an effect on regulations.

Supervisor: Susanne Hindennach

Distribution: 20% Literature, 20% Data Collection/Creation 30% Implementation, 30% Analysis and Evaluation

Requirements: information retrieval, NLP, interest in perception of AI systems (including philosophy/linguistics)

Literature: [1] Gizinski, Stanisław, Michał Kuzba, Bartosz Pielinski, Julian Sienkiewicz, Stanisław Łaniewski, and Przemysław Biecek. “MAIR: Framework for Mining Relationships between Research Articles, Strategies, and Regulations in the Field of Explainable Artificial Intelligence.” arXiv, July 29, 2021. http://arxiv.org/abs/2108.06216.