Towards Inclusive, Contextual, and Balanced Agricultural Advisories: Refining GPT-5 Agricultural Advisories for Kenya with Reinforcement Learning from Human Feedback

This study presents a human-in-the-loop framework to enhance the accuracy, inclusivity, and contextual relevance of GPT-5-based agricultural advisories in Kenya. Using over 2,800 real farmer queries from the iShamba SMS platform, researchers applied prompt optimization, expert review, and Reinforcem...

Descripción completa

Detalles Bibliográficos
Autores principales: Nganga, Kevin Gitau, Grossi, Amanda, Wanjau, Agnes Njambi
Formato: Artículo preliminar
Lenguaje:Inglés
Publicado: Accelerating Impacts of CGIAR Climate Research for Africa 2025
Materias:
Acceso en línea:https://hdl.handle.net/10568/177324
_version_ 1855516185797328896
author Nganga, Kevin Gitau
Grossi, Amanda
Wanjau, Agnes Njambi
author_browse Grossi, Amanda
Nganga, Kevin Gitau
Wanjau, Agnes Njambi
author_facet Nganga, Kevin Gitau
Grossi, Amanda
Wanjau, Agnes Njambi
author_sort Nganga, Kevin Gitau
collection Repository of Agricultural Research Outputs (CGSpace)
description This study presents a human-in-the-loop framework to enhance the accuracy, inclusivity, and contextual relevance of GPT-5-based agricultural advisories in Kenya. Using over 2,800 real farmer queries from the iShamba SMS platform, researchers applied prompt optimization, expert review, and Reinforcement Learning from Human Feedback (RLHF) to refine AI responses. The refined model achieved a 27% increase in satisfactory answers and a technical accuracy score of 1.95/2, outperforming baseline systems. Six key bias types—gender, social, regional, commercial, and linguistic—were identified and mitigated through localized data and bilingual support. The paper demonstrates how RLHF and participatory design can align generative AI with smallholder farmers’ needs, producing advice that is scientifically sound, equitable, and culturally grounded. Findings underscore the potential of inclusive AI frameworks to democratize climate-smart agricultural knowledge while safeguarding against bias in digital extension services
format Artículo preliminar
id CGSpace177324
institution CGIAR Consortium
language Inglés
publishDate 2025
publishDateRange 2025
publishDateSort 2025
publisher Accelerating Impacts of CGIAR Climate Research for Africa
publisherStr Accelerating Impacts of CGIAR Climate Research for Africa
record_format dspace
spelling CGSpace1773242025-11-11T17:09:54Z Towards Inclusive, Contextual, and Balanced Agricultural Advisories: Refining GPT-5 Agricultural Advisories for Kenya with Reinforcement Learning from Human Feedback Nganga, Kevin Gitau Grossi, Amanda Wanjau, Agnes Njambi smallholders-smallholder farmers gender climate-smart agriculture-climate smart agriculture digital agriculture gender equality resilience artificial intelligence climate services-climate information services climate extension agricultural extension This study presents a human-in-the-loop framework to enhance the accuracy, inclusivity, and contextual relevance of GPT-5-based agricultural advisories in Kenya. Using over 2,800 real farmer queries from the iShamba SMS platform, researchers applied prompt optimization, expert review, and Reinforcement Learning from Human Feedback (RLHF) to refine AI responses. The refined model achieved a 27% increase in satisfactory answers and a technical accuracy score of 1.95/2, outperforming baseline systems. Six key bias types—gender, social, regional, commercial, and linguistic—were identified and mitigated through localized data and bilingual support. The paper demonstrates how RLHF and participatory design can align generative AI with smallholder farmers’ needs, producing advice that is scientifically sound, equitable, and culturally grounded. Findings underscore the potential of inclusive AI frameworks to democratize climate-smart agricultural knowledge while safeguarding against bias in digital extension services 2025-09-25 2025-10-24T14:44:03Z 2025-10-24T14:44:03Z Working Paper https://hdl.handle.net/10568/177324 en Open Access application/pdf Accelerating Impacts of CGIAR Climate Research for Africa Nganga K, Grossi A, Wanjau A. 2025. Towards Inclusive, Contextual, and Balanced Agricultural Advisories: Refining GPT-5 Agricultural Advisories for Kenya with Reinforcement Learning from Human Feedback. AICCRA Working Paper. Accelerating Impacts of CGIAR Climate Research for Africa (AICCRA)
spellingShingle smallholders-smallholder farmers
gender
climate-smart agriculture-climate smart agriculture
digital agriculture
gender equality
resilience
artificial intelligence
climate services-climate information services
climate
extension
agricultural extension
Nganga, Kevin Gitau
Grossi, Amanda
Wanjau, Agnes Njambi
Towards Inclusive, Contextual, and Balanced Agricultural Advisories: Refining GPT-5 Agricultural Advisories for Kenya with Reinforcement Learning from Human Feedback
title Towards Inclusive, Contextual, and Balanced Agricultural Advisories: Refining GPT-5 Agricultural Advisories for Kenya with Reinforcement Learning from Human Feedback
title_full Towards Inclusive, Contextual, and Balanced Agricultural Advisories: Refining GPT-5 Agricultural Advisories for Kenya with Reinforcement Learning from Human Feedback
title_fullStr Towards Inclusive, Contextual, and Balanced Agricultural Advisories: Refining GPT-5 Agricultural Advisories for Kenya with Reinforcement Learning from Human Feedback
title_full_unstemmed Towards Inclusive, Contextual, and Balanced Agricultural Advisories: Refining GPT-5 Agricultural Advisories for Kenya with Reinforcement Learning from Human Feedback
title_short Towards Inclusive, Contextual, and Balanced Agricultural Advisories: Refining GPT-5 Agricultural Advisories for Kenya with Reinforcement Learning from Human Feedback
title_sort towards inclusive contextual and balanced agricultural advisories refining gpt 5 agricultural advisories for kenya with reinforcement learning from human feedback
topic smallholders-smallholder farmers
gender
climate-smart agriculture-climate smart agriculture
digital agriculture
gender equality
resilience
artificial intelligence
climate services-climate information services
climate
extension
agricultural extension
url https://hdl.handle.net/10568/177324
work_keys_str_mv AT ngangakevingitau towardsinclusivecontextualandbalancedagriculturaladvisoriesrefininggpt5agriculturaladvisoriesforkenyawithreinforcementlearningfromhumanfeedback
AT grossiamanda towardsinclusivecontextualandbalancedagriculturaladvisoriesrefininggpt5agriculturaladvisoriesforkenyawithreinforcementlearningfromhumanfeedback
AT wanjauagnesnjambi towardsinclusivecontextualandbalancedagriculturaladvisoriesrefininggpt5agriculturaladvisoriesforkenyawithreinforcementlearningfromhumanfeedback