I am a PhD student working on the collaborative content creation use case as part of the UKRI research project PHAWM (Participatory Harm Auditing Workbenches and Methodologies), which is aligned with the UKRI Centre for Doctoral Training in Safe and Trusted AI. My research focuses on the design and evaluation of article generation models using large language models, with a particular emphasis on under-resourced languages. Our goal is to develop models that assist stakeholders, such as Wikipedia article editors, in the article creation process, while also enabling participatory audits of these systems.
More broadly, my research interests span natural language processing and machine learning. For example, while text generated by LLMs has become nearly indistinguishable from human-written text, this advancement also brings significant societal and academic risks, such as misuse and hallucinations. I am also interested in facilitating the development of robust AI detection techniques.
I chose the Safe and Trusted AI CDT because it offers the opportunity to draw inspiration from and learn about a wide range of AI topics through its training and seminars. What stands out most for me is the close collaboration and engagement across cohorts with students from diverse backgrounds and projects, creating a unique environment to nurture and refine ideas.
–
Undergraduate Qualification: BSc Economics, University of Cologne
Postgraduate Qualification: MSc Social Data Science, LSE
Work Experience:
- Research Assistant, Durham University
- Research Fellow, Harvard Kennedy School