Currently, generative AI is not considered capable of creating a viable taxonomy from scratch. The complexity involved in developing a semantically rich taxonomy with appropriate broader and narrower relationships and helpful labels and definitions, free from inconsistencies is still a task that requires human intervention.
Even a fully human-built taxonomy ought to be subjected to continual evaluation and iterative improvements, and one cannot assume that a first draft is fully ready for deployment. That said, taxonomists can and should integrate generative AI strategically and use tools like Copilot to enhance the taxonomy development process while still demonstrating the irreplaceable value of human judgment.
We will examine a hypothetical yet relatable use case and examine how generative AI can act as a helpful assistant at each stage of the taxonomy development process. Defining the scope of the project, identifying existing content, getting valuable insights from stakeholders, drafting early structures and defining categories or facets, fleshing out the taxonomy, and even socializing the taxonomy with future users through card sorts or tree testing can all be assisted by generative AI, albeit cautiously and with oversight.
Bonnie will share an illustrative mix of "weak prompts" and "strong prompts" and evaluate their strengths and shortcomings in the various stages of taxonomy development. Delegates can expect these examples to range from the simple and straightforward (proposing alternative labels and definitions) to the more complex (working through card sorts and tree testing) and will leave ready to test out new prompts in their own work.
