This project aims to develop an AI-driven framework that learns from experts, works alongside them, and gets better the more it is used. The core idea is a human-in-the-loop learning system in which expert reviewers actively shape the model rather than simply receiving its outputs.
Using reinforcement learning from human feedback (RLHF), the system updates its behaviour based on how experts accept, correct, or reject its assessments, gradually learning the reasoning patterns of experienced practitioners rather than just statistical patterns in data.
Over time, the model becomes calibrated to the standards and priorities of the expert community using it.
The system can draw on a range of data types, including satellite and drone imagery, underwater acoustic recordings, species survey data, or regulatory documents, processed through a vision-language modelling framework with a conversational interface that allows experts to interrogate outputs, flag uncertainty, and provide targeted corrections.
The specific data modality or combination that the successful student pursues will be agreed based on their background and preferences following recruitment.
The project aims to develop an AI system that accelerates EIA without replacing the human judgment that makes assessments credible, legally defensible, and ecologically meaningful. It can be validated across contrasting UK or European OSW sites, contributing both novel methodology and open tools for the broader EIA community.
The programme includes a structured training component delivered through compressed teaching weeks during the six-months, with CDT funding covering travel costs for students based at partner universities.( Plus any genuine training components that your PhD project entails.)
94% of Loughborough’s research impact is rated world-leading or internationally excellent. REF 2021