Mahdis Tourian

Environmental Intelligence: Multi-agent evidence synthesis for complex environmental data

Completion: 01 SEP 2028

Background

I come from industrial engineering and hold an MBA. I am motivated by real-world decision-making and systems thinking, optimising limited resources, quantifying uncertainty, and designing user-centred tools. Those strengths translate directly to environmental intelligence: structuring evidence, building reliable AI pipelines, and supporting transparent trade-offs in nature recovery planning.

PhD

Environmental Intelligence: Multi-agent evidence synthesis for complex environmental data

My PhD explores how artificial intelligence can support environmental decision-making by helping experts bring together fragmented ecological evidence from multiple sources, such as habitat data, species records, environmental datasets, guidance documents, and reports.

The project investigates the development of a multi-agent evidence synthesis tool that can retrieve relevant information, assess data quality, combine evidence, and generate structured, transparent outputs for environmental reporting workflows. Rather than replacing expert judgement, the system is designed to support practitioners by making evidence synthesis more consistent, traceable, and easier to review.

The research is currently exploring potential use cases including habitat assessment, species monitoring, environmental assessment, and policy evidence synthesis. A key focus is responsible AI: ensuring outputs are linked to sources, uncertainty is clearly shown, and humans remain in control of decisions.

Methods include multi-agent system design, retrieval-augmented generation, environmental data integration, uncertainty representation, and human-AI interaction evaluation with environmental practitioners.

In Simple Terms...

Environmental decisions often depend on many different types of evidence, including maps, species records, reports, and guidance documents. My research asks whether AI can help bring this information together in a clear and trustworthy way.

I am developing ideas for a tool that acts less like a general chatbot and more like a team of specialist assistants. One part finds relevant evidence, another checks its quality, another combines the information, and another helps produce a structured draft report.

The goal is not to let AI make environmental decisions. The goal is to help experts save time, see where evidence comes from, understand uncertainty and gaps, requirements for fieldworks, and produce more transparent reports that support conservation, planning, and nature recovery.

Supervisors

Lead Supervisor
Professor Hywel Williams

Computer Science · Faculty of Environment, Science and Economy (ESE)
Co-Supervisor
Dr. Sareh Rowlands

Computer Science · Faculty of Environment, Science and Economy (ESE)

External Partners

Natural England
Natural Resources Wales

Back