Research Scientist for Behavioural Science and AI Research - AI Safety Institute

About The Job

The AI Safety Institute (AISI) is looking for a Research Scientist with expertise in both Behavioural Science and AI research to help lead studies that explore the impact of advanced AI models on humans. These experiments are a core part of our research, which is geared towards assessing the capabilities and risks of advanced AI systems. We wish to study how humans use AI systems, how their capabilities are enhanced by AI systems, and how human users themselves may be psychologically affected by using AI systems. The post-holder will help lead / coordinate research in this area, by designing experiments, training and evaluating AI models (especially large language models [LLMs]), analysing and modelling human data, and summarising results for internal and external dissemination.

The AI Safety Institute is the first state-backed organisation focused on advanced AI safety for the public interest. We launched at the AI Safety Summit because we believe taking responsible action on this extraordinary technology requires a capable and empowered group of technical experts within government. Our staff includes senior alumni from OpenAI, Google DeepMind, start-ups and the UK government, and ML professors from Oxford and Cambridge. We are now calling on the world’s top technical talent to build the institute from the ground up. This is a truly unique opportunity to help shape AI safety at an international level.

We have ambitious goals and need to move fast. Our top priorities are to:

• Develop and conduct evaluations on advanced AI systems. We will characterise safety-relevant capabilities, understand the safety and security of systems, and assess their societal impacts.
• Drive foundational AI safety research. We will launch moonshot research projects and convene world-class external researchers
• Facilitate information exchange. We will establish clear information-sharing channels between the Institute and other national and international actors. These include policymakers, international partners, private companies, academia, civil society, and the broader public.

Job Description

We are looking for a Research Scientist with training in behavioural science, and experience of conducting research in machine learning / AI. The ideal candidate will have a vision for how to run studies that assess the impact of AI systems on humans, in order to carry out high-impact research in AI safety.

They will have experience with collecting and analysing human behavioural data, ideally in a setting where humans interact with AI systems. They will have experience with working with advanced AI models, for example in conducting research that studies the impact of prompting or fine-tuning LLMs on relevant outcomes. They will use their skills and experience to conduct research that measures the positive and negative impacts of AI systems on individuals and groups.

These could be psychological impacts, related to mental health (e.g. does interacting with models change mood or levels of anxiety) as well as epistemic health (are people more likely to believe fake news or subscribe to conspiracy theories after interacting with the models). We are also very interested in how advanced AI systems may enhance human capabilities, especially in domains where this uplift may have dangerous consequences (e.g. biosecurity or cybersecurity).

Person specification

The ideal candidate will have excellent research skills, experience with designing human behavioural experiments and collecting and analysing human data, and AI research skills, including working with advanced AI models (such as LLMs).

Example duties on any given day could include:

• Designing a new human-AI study
• Providing guidance to other team members responsible for building experimental UI
• Creating an analysis pipeline for large-scale human data
• Monitoring study progress via reports from other team members
• Prompting or fine-tuning an advanced AI system to make it safer or more capable
• Writing research papers or delivering talks that describe research findings

The ideal candidate will have skills and experience that allow them to:

• Lead on design of high-quality research that evaluates advanced AI systems with a focus on safety
• Design human-AI interaction studies rigorously, ethically, safely, with both data quality and participant experience / wellbeing in mind
• Identify improvements to research workflows and troubleshoot problems as they arise
• Design analytic pipelines that address key research questions, and provide technical oversight of their implementation
• Serve as a mentor or line manager to more junior colleagues
• Liaise with the team lead and wider AISI to respond to goals and to deliver impactful research

Essential requirements for the role this include:

• Significant experience of leading experiments that involve human data collection, for example from graduate-level study in cognitive or social sciences, or (extensive) equivalent experience in an industry setting
• Good technical and coding skills
• Experience of mentoring others in a research setting
• A significant track record of publishing work, disseminating findings, or other pathways to impact
• Experience with conducting research that involves online collection of human behavioural data (e.g. Prolific Academic, Amazon Mechanical Turk, or an equivalent platform)
• An understanding of the ethical and data protection issues associated with human experimentation (e.g. Research Ethics, GDPR)
• Strong project management skills and demonstrable experience of teamwork, including building rapport with people from diverse backgrounds