Towards GilGPT: an anthropomorphic conversational agent for accessible communication of agricultural research to farmers

The use of generative AI-based conversational agents is becoming increasingly prevalent in the dissemination of knowledge. In the agricultural sector, there is a substantial corpus of scientific literature, yet there is frequently a deficit of accessible syntheses that can effectively disseminate this knowledge. To enhance the accessibility of scientific findings for farmers and researchers, we introduce our anthropomorphically designed Large Language Model based conversational agent, “GilGPT”. Developed as a design science research project, it uses Retrieval Augmented Generation to access scientific literature from a knowledge base. This approach enables a natural language dialogue between scientific publications and farmers. The implementation of an anthropomorphic design is achieved through the utilization of avatars and adaptive language among other techniques. This approach is employed with the objective of enhancing the accessibility of complex scientific content and facilitating the dialogue between research and practice.