Evaluating the Effectiveness of Large Language Models for Livestock and Climate-Related Agricultural Advice in Ontario
Bibliographic record
Abstract
Recent advancements in artificial intelligence (AI) and large language models (LLMs) offer new opportunities for improving agricultural extension, particularly in communicating livestock and climate-related knowledge. While general-purpose models like ChatGPT have demonstrated potential, their performance in specialized domains such as animal welfare has yet to be fully assessed. Prior studies suggest that domain-specific models outperform general ones on precision and contextual accuracy, yet comparative evaluations with expert-curated content are limited. This study examines the performance of ChatGPT, Claude, and Gemini in answering livestock-related questions relevant to climate change. It evaluates the degree of alignment between AI-generated and expert-developed answers, focusing on five metrics: similarity, faithfulness, context precision, context recall, and answer relevancy. Ten expert-reviewed questions were developed, and corresponding human-curated answers were constructed from recent literature. Responses from the three AI models were collected using standardized prompts. Answers were evaluated using a multi-criteria framework supported by qualitative coding and statistical summaries. Prompt engineering was applied to improve answer quality and comparability across models. AI models—especially ChatGPT and Claude—showed high alignment with expert answers. Their outputs demonstrated strong similarity, faithfulness, and context relevance. While some variation in depth and specificity remained, the overall quality of AI responses was high across most metrics. LLMs show promise for supporting agricultural extension and public knowledge transfer. Ensuring reliability requires continued use of expert oversight, domain-specific data, and refined prompting strategies.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.007 | 0.042 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.003 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".