Like many others in law schools and beyond, I have been experimenting with Chat GPT to better understand its capabilities and limitations. From an online seminar led by Mushtaq Bilal (Literature, University of Southern Denmark), I learned that about customizing ChatGPT (clicking on your name in the lower left corner of the home screen-->click "Customize ChatGPT"). You can provide information about your areas of academic interest and how you want ChatGPT to respond (i.e., use a formal, academic tone; here's what I expect you to address if I ask for a critique of a paragraph). My anecdotal experience so far is that providing these customized instruction makes ChatGPT a more accurate source of assistance.
Separate and apart from availing oneself of the "Customize ChatGPT" screen, one can also ask ChatGPT to remember certain things, including what you have written in the past. That way, ChatGPT has a better sense of the tone, output, etc. you may be expecting (more explanation here).
I recently experimented by uploading to ChatGPT four of my published law review articles and asking it to remember that I am the author of those works. I then asked ChatGPT to "write a paragraph about balloons in the style of Bridget J. Crawford." Here were the results:
I have never written about balloons before, and I haven't written anything substantial about environmental issues (or anything about consumer safety). So it is difficult to say if this is written in my "style." But I do think that ChatGPT's output in response to this prompt has some noteworthy qualities that seem "law prof-ish" to me: a coherent thesis statement, a mention of variations in legal framework, the reference to regulations, and linking to other issues ("a microcosm for examining law legal systems can evolve"). I'm interested in continuing to experiment--cautiously, ethically--with this technology.
Has anyone else experimented with custom instructions? Any tips for what is working well (or poorly) at this stage?
Comments