AI for Science and Engineering Advice

Project ID: 2228cd1322 (You will need this ID for your application)

Research Theme: Artificial Intelligence and Robotics

UCL Lead department: Science, Technology, Engineering and Public Policy (STEAPP)

Department Website

Lead Supervisor: Chris Tyler

Project Summary:

This project would explore how artificial intelligence (AI) could and should be used to enhance the practice of translating scientific and engineering knowledge into public policy domains (‘science advice’). AI technologies have the potential to change the way that science advice is done in four key areas: synthesising evidence, horizon scanning, drafting policy briefs, and impact evaluation.

Science advice has been done the same way for decades. Advisers’ approaches to evidence syntheses and horizon scanning tend to be rather ad hoc. Policy briefs are one-size-fits-all and take months to produce. Impact evaluation often amounts to counting publications and event attendees. AI tools could revolutionise all these areas, from automating parts of the evidence synthesis process, to producing multiple versions of personalised policy briefs.

The design and application of future AI tools in science advisory systems will inevitably be complicated by legitimate concerns about rigour and trustworthiness. These tools must be co-created with academia and science advisory institutions. Processes will need to be developed to deal with system glitches, such as the tendency for ChatGPT to show political bias.

This PhD project would have two main goals. The first would be to identify AI technologies for science advice, and an assessment of opportunities and challenges in their use. The second goal would be to deliver real-world impact in policy institutions. The programme would be designed from the outset as practice-oriented and would aim to offer advice on and toolkits for the deployment of AI technology for science advice, and/or the production of AI tools for use in science advice.

We are looking for a PhD candidate with a background in computer science (essential) and experience in public policy (desirable).