The global south is particularly vulnerable to the impacts of climate change, such as extreme weather events, sea level rise, and changes in temperature and precipitation. It is therefore critical that agricultural systems are made resilient to climate, so that food production can withstand, recover from, and adapt to climate variability and extremes. One way of achieving this is through new technologies and practices to improve resilience by maintaining crop yields whilst reducing environmental impact and enhancing the sustainability of agriculture. These agronomic innovations need to be carefully selected from the many available, and targeted towards the different types of farms that exist. LLMs (Large Language Models) have the potential to address this challenge by synthesising and communicating information in a way that is tailored to the needs of different groups of farmers.
This studentship uses recent progress and the resources of a newly-funded project, iSPARK, to develop new analyses and LLMs that can select and disseminate appropriate agricultural innovations in a timely and relevant way through agronomic advisory services. iSPARK is an international collaboration between various CGIAR research centres and demand partners that is developing and implementing sustainable and climate-smart agricultural innovations in western Kenya. There are two key elements of iSPARK that will directly feed into the studentship:
1. Farm typologies are classifications of farms based on their shared characteristics, such as size, production type, management practices, and ownership structure. they are useful for understanding the diversity of farms and for targeting agricultural policies and interventions to the specific needs of different groups of farmers. iSPARK is building on and developing farm typologies for Western Kenya.
2. iSPARK is developing a database of agronomic innovations focussed on climate resilience. It contains, for example, AgWise a generalizable decisions support framework that has been used to develop initial sub-county level fertilizer recommendations that have the potential to increase maize yields by 30% over current averages in western Kenya.
LLMs can be trained on a variety of data sources, including the database of agronomic innovations, to generate text, translate languages, write different kinds of creative content, and answer questions in an informative way. The project will draw on early progress in this field, much of it made by one of the supervisors.
Research Question, hypothesis, and Objectives
The student will address the question: how can LLMs be developed to communicate from the database of agronomic innovations in a way tailored to the iSPARK typology? The underlying hypothesis is that tailored LLMs can lead to improved adoption of existing agricultural innovations and thus higher and more resilience crop productivity.
The objectives of this PhD project are to:
1. Adapt and train LLMs to synthesise, target and communicate information from the iSPARK database to directly address farmers’ needs. Understanding of farmers’ needs will be based on the iSPARK typologies and on the cultural, economic, and social factors identified by our on-the-ground partners.
2. Evaluate the effectiveness of the developed LLMs in communicating information to different groups of farmers using defined metrics such as comprehension, recall, and actionable insights.
3. Provide recommendations for optimising communication of LLMs based on evaluation results and feedback from the target audience.
Supervision, project partners and associated opportunities
The PhD will have four supervisors. Andy Challinor, in the School of Earth and Environment, is Professor of Climate Impacts and focussed mainly on food systems. Chetan Deva is a senior postdoctoral researcher in Andy’s research group. Anthony (Tony) Cohn is Professor Automated Reasoning in the School of Computing and has jointly supervised students with Andy over recent years. Aniruddha Ghosh is a Senior Scientist at Alliance Bioversity-CIAT with significant experience in developing LLMs for agronomic advisory services.
The project will provide exciting collaboration opportunities with globally recognized agricultural research institutions within the CGIAR, such as the Alliance of Bioversity and CIAT. The Alliance addresses urgent global challenges like malnutrition and climate change through cutting-edge research. PhD students not only contribute to transformative studies but also benefit from the Alliance's global presence in over 20 countries, enabling collaborations with worldwide experts.