Postdoctoral Research Associate in NLP, LLM Uncertainty Expression, and AI Ethics
King's College London is recruiting a
Postdoctoral Research Associate
for work at the intersection of
Natural Language Processing
,
Large Language Models
,
AI ethics
, and
human reasoning
.
The role sits within the
Centre for Data Futures
and the
Dickson Poon School of Law
, in collaboration with
Prof. Yulan He
and her NLP group at King's Institute for Artificial Intelligence. The project investigates how different modes of AI uncertainty expression affect the quality of human reasoning and deliberation, especially in ethically loaded contexts such as healthcare, law, and education.
The successful candidate will lead the design, implementation, and evaluation of computational experiments using LLM-based systems. Core tasks include building experimental interfaces, developing prompting and fine-tuning strategies to vary uncertainty expression, and creating evaluation pipelines to measure downstream effects on human participants. The research is framed around concepts such as
doxastic plasticity
, deliberative openness, and collective norm-refinement.
Eligibility highlights:
a PhD in Computer Science with NLP expertise is essential, or a thesis submitted with viva pending. Strong experience with LLMs, fine-tuning, prompt engineering, and evaluation is expected. A track record of peer-reviewed research is required, and interest in social science, humanities, AI ethics, or human-computer interaction is valued. Experience with human participant studies, participatory methods, open datasets, or policy/practitioner communication is desirable.
Funding and terms:
this is a full-time 24-month postdoctoral contract, with possible extension depending on further funding. Salary is £45,031-£47,379 per annum including London Weighting Allowance. The project is supported by The Patrick J. McGovern Foundation.
Deadline:
17 June 2026.
How to apply:
submit a CV and supporting statement through the King's College London job portal. The advert asks applicants to address the essential criteria and, where possible, the desirable criteria.
