About the Workbench

A site about data should model good data practices

This page says plainly what this site is, who it is for, what it collects, and how it is built — because reading a clear data notice should be the first artifact of analysis for anyone studying these systems.

What this is, and who it is for

The Workbench is a set of interactive labs on AI and legal practice, built alongside the law school course Law Practice in the Age of AI & Big Data and free for anyone to use — students, practitioners, judges, librarians, the curious. The labs are organized by the four roles a lawyer occupies around AI: the User operating the tool, the Evaluator deciding whether it is fit for purpose, the Builder creating or commissioning a system, and the Steward shaping how AI enters the legal system itself. Start with the orientation lab, The Four Roles, and go from there.

Every lab passes one test before it earns a place here: if you could have the experience with a legal research database or a free chatbot, it does not belong. The labs exist to show what those tools cannot show you — the machinery underneath the chat window, the decisions baked into training data and system design, and your own instincts, measured against your actual accuracy.

The labs are ungraded on purpose. They are a place to practice with the ideas, at your own pace, without an audience.

What this site collects: nothing

This is a static site. It has no server-side code, no accounts, no analytics, no cookies, no trackers, and no third-party requests — the fonts are self-hosted, so not even a font service sees that you visited. Your clicks, answers, confidence ratings, and results exist only in your browser's memory while a lab is open. Close the tab and they are gone.

A few future labs may offer to save your work on your own device (in your browser's local storage) so you can pick up where you left off. Any lab that does so says so on its face, in plain language, with a one-click way to erase everything. Nothing is ever transmitted anywhere, to anyone, under any circumstances.

That is the whole policy. If a data-use notice can fit in three short paragraphs, it should.

Accessibility

Every lab is built to be operable by keyboard alone, to work with screen readers (real form controls, visible focus, live-region announcements for dynamic feedback), to respect your reduced-motion settings, to never rely on color alone to convey meaning, and to run on a mid-range phone — not just a new laptop. Every chart ships with its data in an ordinary table. Tools about exclusion cannot themselves exclude; if a lab falls short of that standard for you, that is a bug, and reporting it is a favor.

A note on the fabricated citations

Some labs deliberately contain invented cases and misattributed holdings — that is what they teach. Every fabricated authority is revealed and corrected inside the lab itself, and nothing on this site should be cited for anything, ever, without independent verification. A site that teaches verification should say so about itself.

Use of AI in building this site

Generative AI helped build this site. Our course policy is verification and disclosure, not prohibition, and that policy applies to me first: AI drafted code and prose from my frameworks, lab designs, and writing samples, and I directed, revised, and verified everything before it shipped. Every fabricated authority is deliberate and labeled inside its lab; any error beyond those is mine, and I would genuinely like to hear about it. A site constantly in beta can be fixed. For the long-form version of this kind of disclosure, including estimated environmental costs, see the AI statement in my open legal research textbook.

Credits

Built by Cas Laskowski. Archivo, IBM Plex Mono, and Newsreader are used under the SIL Open Font License.