Human–AI workflows assessed in California vineyards
A precision disease-control project examined the value of two live human–AI collaborations in California vineyards. Forecasts produced through one workflow were compared with independent field scouting conducted in 2025.
- Source
- arXiv – Viticulture & Grapevine
- Published
- Reading time
- 2 min read
- Region
- International
Data to support prioritized scouting
A California precision disease-control project assessed the value of two live interactions between people and artificial intelligence. The study examined whether commercial scouting records collected from 2021 through 2024, together with remote-sensing measurements, could be used to forecast red-leaf symptoms in 2025.
The work covered 140 hectares of vineyard area. Its practical objective was to support the prioritization of field scouting and virus testing through symptom forecasts. The material provided does not identify the vineyards, describe the remote-sensing instruments or specify which viruses were included in the testing effort.
Iterative human refinement
In the first workflow, a multi-agent research system called Aleks v1 developed forecasting models with iterative human refinement. The approach therefore did not rely on a fully automated model-development process: human input formed part of the research cycle as the models were developed and adjusted.
The researchers applied Aleks’s 2024 vine-scale model to updated predictors for 2025. They then evaluated the resulting red-leaf forecasts against independent scouting performed during 2025. This design enabled a model derived from earlier information to be compared with observations collected in a subsequent season.
Limits of the available details
According to the title and excerpt, the broader project evaluated two human–AI workflows, but the supplied material describes only the first one. It does not explain the design of the second workflow, identify the evaluation metrics or provide numerical results for model performance.
The available information therefore does not establish how accurately the system forecast red-leaf symptoms or how much it improved the prioritization of scouting and virus testing. No comparison between the two workflows can be made from the supplied details either. What the study does document is a field-based use case combining several years of commercial records, remote-sensing measurements, artificial intelligence and iterative human guidance at vineyard scale.
The original study was published by arXiv – Viticulture & Grapevine.
Related articles
More stories from the same topic area.
Back to news
TechnologyDeep Learning Identifies Leafhopper Damage in Vineyard Conditions
Research published in Smart Agricultural Technology reports that deep learning models can reliably detect characteristic leaf damage caused by grapevine leafhoppers under real vineyard conditions. The available summary does not provide details about the models, dataset or accuracy results.
Bing News – Vineyard Technology · 2 October 2026 · 2 min read
TechnologyVineyard Drone Models May Lose Reliability Across Seasons
A summary published on September 2 says that the performance of drone models used in vineyards may not remain stable from one season to another. The available source does not provide details about the models, data or evaluation metrics examined.
Google News – Viticulture Robotics & Sensors · 26 September 2026 · 2 min read
TechnologyStudy Tests Two Satellite Models for Vineyard Water Use in Bekaa Valley
A study examines two satellite models in relation to vineyard water use in Lebanon’s Bekaa Valley. The available source does not identify the models or report their methodology and performance.
Google News – Vineyard Computer Vision · 24 September 2026 · 2 min read