Research Scientist – Human Influence (AI)
Мэтч & Сопровод
Для мэтча с этой вакансией нужен Plus
Описание вакансии
Location: London, UK, with hybrid working and options for similar government offices in Birmingham, Cardiff, Darlington, Edinburgh, Salford or Bristol; occasional remote work abroad may be possible.
Annual salary: £65,000–£145,000, comprising base salary and technical allowance, plus a 28.97% employer pension contribution on base salary.
Company
The is a UK government research organisation focused on understanding advanced AI risks and translating research into action with governments and frontier AI developers.
What you will do
- Lead and contribute to research programmes on how frontier AI systems influence people and other AI agents.
- Design, plan, conduct and analyse rigorous human-AI interaction studies, including large-scale experiments and randomised controlled trials.
- Build and run AI model evaluations, benchmarks, red-teaming exercises and datasets tracking real-world risk exposure and severity.
- Apply computational social science, cognitive and behavioural science, machine learning, data science and AI safety methods.
- Publish and disseminate research through academic papers, preprints, blog posts, talks and posters.
- Engage with government officials, policymakers, regulators, frontier lab employees and the wider research community.
Requirements
- PhD or equivalent years of research experience in a relevant field.
- Track record of productive, rigorous and high-impact research publication.
- Experience designing, planning, conducting or analysing human experiments, ideally online and at scale.
- Strong knowledge of advanced statistical modelling, including multilevel, hierarchical or mixed-effects GLMs.
- Knowledge of frontier AI models and how they are trained and evaluated.
- Strong coding skills in Python and/or R, including experience using AI coding tools such as Claude Code, Codex or similar.
- Experience with AI model or system evaluations and other AI safety projects.
- Strong verbal and written communication skills and demonstrable interest in the human impacts of AI.
- Applicants must complete baseline security checks, including a DBS check; SC clearance is not required for this role.
Nice to have
- Published work on societal impacts of AI, AI system evaluation, multi-agent systems or a related field.
- Experience fine-tuning language models or working with reinforcement learning.
- Experience analysing large-scale multi-turn conversations or chat logs using LLM judges or auto graders.
- Front-end software engineering skills for building study interfaces for human participants.
Culture & Benefits
- Work on high-stakes, policy-relevant AI safety research with access to pre-release frontier models, large compute budgets, API credits and funding for human studies.
- Opportunities to lead research, publish externally, collaborate with experts across national security, policy and AI research, and influence frontier AI governance.
- Five days off and annual learning and development stipends, with funding for conferences and external collaborations.
- Hybrid working, occasional remote work abroad, at least 25 days of annual leave, eight public holidays, team-wide breaks and three volunteering days.
- Generous paid parental leave, pension contributions, work-from-home equipment support and other practical benefits.
Hiring process
- The application deadline is Sunday 29 November 2026, end of day, anywhere on Earth.
- Applicants should provide truthful, factually accurate examples based on their own experience; plagiarism or presenting AI-generated work as personal experience may result in withdrawal of the application.
- Successful candidates must complete pre-employment checks and BPSS clearance before appointment; eligibility for CTC clearance is preferred for some work.
Будьте осторожны
Если работодатель просит войти в их систему, используя iCloud/Google, прислать код/пароль, запустить код/ПО, не делайте этого. Это мошенники. Обязательно жмите «Пожаловаться» или пишите в поддержку. Подробнее в гайде →