In my earlier post, I attempted to “assess” CG the way I might assess a student. This assessment led me to the uncomfortable conclusion, that, in this case, it was indeed a very good, albeit uninteresting, student. It was the kind who tells me exactly what I just taught them and then loop this back to myself. It was a perfect artificial student and I do not need any of those.
I closed the post with a question to open a different line of investigation. Shoud I treat ChatGPT (CG from now on, as type of a social fact often glossed as “AI”) as an “educator”? Or, at least, in Cremin’s phrase, an “educational institution.” The answer has to be “Perhaps…. but…”
What kind of educator might CG be? To answer this, an anthropologist wonders about what it is actually used for. By every account it is used for a lot of stuff, including some stuff many people think it should not be used to, or should be prevented from doing. I have used it most successfully for technical questions about programming details. I know advanced programmers use it extensively for this. I have collected stories from acquaintances about their use of CG. I watched generated videos on YouTube. I know of a young close relative of mine, who used it explore, while confined at home during “Covid,” marxism-leninism and the history of Russia, as well as neo-naziism around the world—with no guidance or control from any other person. Many use it to check their written English. And everyone, from government officials, to journalists, to everyone with whom I talked can be produce extended discourses about the good or bad of CG. It is undoubtedly a “total social fact” no one can escape.
The part of me who teaches Rancière to make students aware of all aspects of education that escapes the State can only celebrate all this. CG might be the ultimate “ignorant school master”—except that it is one that knows everything! So, CG is more like an infinite library open day and night, 24/7 to anyone interested to browse through it. It is a library with an altogether kind librarian that keeps close track of what one has explored and responds to further investigations with an encouraging “this a good question, let me think” before summarizing what one might have read and, perhaps, send one into a further investigation deepening the initial one.
But CG is not at all like the, Jacotot, the schoolmaster who inspired Rancière. This schoolmaster may be ignorant of the topic interesting to a student but he is actively involved in prodding their will to continue searching and reaching a new mastery. By contrast, so far at least, CG is silent until one asks. This silence is actually, for me, just what I want from a library, and an assistant librarian. I use CG regularly because it is indeed a a powerful tool for an intellectual, a major advance over Google.
But this silence or relative passivity also suggests that CG is not an “educator” in any of the usual senses. It is not a teacher. It is not a master. It is not a mentor. It does not have a curriculum. It does not mete consequences.
To push this, I asked ChatGPT “what should I know about the creationism museum”? Having visited it, I can say that answer was clear, succinct, as well as expanding on the controversies surrounding it. And CG ended as usual, with suggestions for further explorations: “If you want, I can also summarize how this museum is seen by scientific organizations or provide tips for visiting (hours, ticketing, best time to go).” There were links to further sources.for the theoretically inclined this is an instance of the “instruction writer” limits sketched by Garfinkel (2002: Chapter 6).
But there was no clear indication that using the answers “naively,” as one might use other answers, might land one in very dangerous interactional or political settings. CG, so far, does not (cannot?) know the social contexts within which its answers might be used.
“We” (teachers, professors, educators…) must, at least, warn our students…