Development of an Exploratory Taxonomy for Veterinary Professionals' AI Query Patterns Across Clinical Stages: An Expert Panel Study
This study establishes the first veterinary-specific, clinical-stage-sensitive taxonomy of AI query patterns by analyzing real-world chatbot logs and refining them through an expert panel review, resulting in a framework of three categories and 21 subtypes to guide the development of context-aware veterinary AI systems.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer
Imagine a veterinary clinic as a busy kitchen where chefs (vets) are constantly cooking up solutions for sick animals. Recently, they've been given a new, super-smart sous-chef: an AI chatbot that can answer almost any question. But until now, nobody had a map of what questions the chefs were actually asking this new helper.
This study is like a team of master chefs and food critics getting together to organize a massive pile of 5,372 notes left by vets on how they used this AI. Their goal was to sort these notes into a neat, easy-to-read menu—a "taxonomy"—that shows exactly how vets interact with AI at different points in treating an animal.
Here is how they did it and what they found:
The Recipe for Discovery
First, the researchers used a computer to read through thousands of real questions vets asked over eight months, grouping similar ones together like sorting ingredients into bins. Then, they invited a panel of 38 experienced veterinary experts to taste-test this initial list. Through a structured survey, these experts refined the list, making sure it made sense in the real world.
The Three Main Courses
The final result is a menu with three main sections (categories) containing 21 specific types of questions:
Clinical Support (The "What's Wrong?" Course): This is where vets ask the AI to help figure out what an animal might have or how to treat it.
- The Star Dish: The most popular question type was "Differential Reasoning" (Type B). Think of this as the vet asking, "If a dog has these symptoms, could it be A, B, or C?" This was the most common question overall.
- The Timing: Right after seeing a patient, vets mostly asked for "Clinical Decision Support" (Type D), like asking, "Okay, I've seen the patient; what's the best next step?"
Evidence-Based Research (The "Show Me the Proof" Course): This is for vets who want to dig deep into scientific studies to back up their choices.
- The Special Ingredient: Vets with more than 10 years of experience asked for "Evidence Search" (Type G) much more often than newer vets. It's like a veteran chef double-checking a classic recipe against a new book, while a junior chef might just trust the recipe as is.
- The University Crowd: Vets working at universities were almost obsessed with this "Evidence Search" type, making it their go-to question.
Terminology and Drug Reference (The "Dictionary" Course): This covers simple look-ups for hard words or checking drug dosages.
Why This Matters
The paper claims that no one has ever built this specific "menu" for veterinary AI before. By understanding exactly what questions vets ask and when they ask them (like during the diagnosis vs. after the consultation), the study provides a blueprint. This blueprint is meant to help designers build AI tools that know exactly what a vet needs at every specific stage of a patient's care, rather than just being a generic question-answering machine.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.