Plant growth-promoting bacteria and generative AI validation in sustainable lettuce seedling production systems.

FONTENELLE, M. R.; SILVA, K. M. N. B. da; SILVA, K. C. R.; RODRIGUES, G. S. S.; JORGE, M. H. A.; LIMA, C. E. P.

Plant growth-promoting bacteria and generative AI validation in sustainable lettuce seedling production systems.

Autores

FONTENELLE, M. R.

SILVA, K. M. N. B. da

Descripción

The production of lettuce seedlings (Lactuca sativa L.) is a fundamental process for maintaining the good quality of the final product, as well as for maintaining the sustainability of crops. The objective of this study was to evaluate the potential of nine bacterial isolates to promote the growth of lettuce seedlings. To this end, two production cycles were conducted using 120-cell trays and inert substrate fertilized with mineral fertilizers in a greenhouse at Embrapa Hortaliças, Federal District, Brazil. A completely randomized design (CRD) was used with five replicates and 12 treatments, including three controls (use of mineral phosphorus, no use of mineral phosphorus, and use of a commercial P-solubilizing biofertilizer). The other nine treatments consisted of bacterial inoculants with phosphate solubilization potential and siderophore production previously tested in vitro. In these, no mineral phosphorus was added during fertilization. The following variables were evaluated: average number of leaves, average leaf width, aerial part length, and root length. The data were submitted to analysis of variance and the means were grouped by the Scott-Knott test. The quantitative consolidation of the results of the two production cycles was done by calculating the average rank. Visual analysis was also conducted by human experts (two PhDs) and Generative Artificial Intelligence (AIs - LLMs). For this purpose, ChatGPT5, Gemini, and Copilot were tested. The analysis was performed using Prompt Engineering and Prompt Chaining proposed in this work and photos attached to the AIs. Only ChatGPT and Gemini showed good results, compatible with human observation (appropriate Spearman's Correlation Coefficient and Mean Deviation Average). The consolidation of the rankings based on quantitative data and visual analysis by the AIs and human observation identified T7, T5, and T8 as the best treatments, with T7 standing out. This work is in line with the current needs of sustainable, resilient, and adapted agriculture.