AI Workbench collects experiments with AI in humanities research at UNM. Each account starts with a particular task: transcribing a manuscript, getting references into Zotero, locating a photograph, checking a draft, or building a model. We describe what we tried, what came out, and what still needed a researcher’s judgment.
We call each experiment a jig. In a woodshop, a jig is a guide built for one job: it holds the work and guides the tool so the same cut can be made again. A jig here does that for one research task, recording the sources, tools, prompts, and steps that went into it.
A useful output and a reliable source are different things. The workbench makes room for partial results, failed attempts, and decisions to leave AI out of a workflow. The point is to give colleagues enough detail to try an approach and assess its limits.
For classroom activities and assignments, visit AI Sketchbook.
Email us a few paragraphs about a research task, the tools and sources you used, and what you checked or corrected. An unresolved problem is welcome too.
The contribution guide explains the format.