First, I was surprised to learn that most of them did not know anything about ChatGPT. Nothing. One lone person raised their hand when I asked who had heard about it. No one had used it.

I'm sure there are a lot of reasons for this, but clearly this is more on the brains of faculty and administrators than it is even a blip on the radars of my students. 2/11

Here's my prompt and its response:

My Q: Write a one paragraph summary of this document: Richard Hakluyt Makes the Case for English Colonization, 1584

ChatGPT:
In "Richard Hakluyt Makes the Case for English Colonization, 1584," the author, Richard Hakluyt, presents a strong argument for the colonization of North America by the English. He cites the benefits of such an endeavor, including the potential for trade and economic gain, the spread of Christianity, and the potential for finding new resources and lands. Hakluyt also addresses potential objections to colonization, such as the cost and potential danger, and counters them with reasons why the benefits outweigh the risks. He concludes by urging the English government and people to take action and pursue colonization in the interest of the country's prosperity and glory. 3/11

Students were actually pretty observant as they assessed this summary. One noted that the summary makes the source feel much more balanced than it is, as though Hakluyt gives equal time to objections as he does to his arguments for colonization. I asked students to think about what a more honest portrayal of Hakluyt's proportions would be, which opened up a discussion on the piece as a fundamentally persuasive text.

I think there was a missed opportunity here on my part to talk about why ChatGPT might have chosen to interpret the source in this way. I've seen in other questions I've posed how ChatGPT really visibly strives toward some sort of neutral, objective perspective in just about every request. Here, you can see that tendency as one that actually distorts the nature of the material in question. There's a concept that is difficult to get at in the classroom that was easier to touch on here, even if just for a moment. 4/11

We agreed as a group that nothing was factually wrong with this account but, when asked what was missing, students were quick to identify a huge gap: there is no mention of Spain. The original document mentions Spain directly or indirectly constantly. ChatGPT didn't include it at all.

I've been working really hard to frame early colonial history as playing out against a dense and tangled web of geopolitical relationships: indigenous groups wrestling with individual European groups and vice-versa, as well as European powers vying for dominance with other European powers.

This exercise encouraged the students to identify that missing piece on their own (ok, with a tiny bit of prompting from me) and thus underscored, I think, the larger theme of the session and the last few classes. That was a big win and took very little effort. 5/11

This omission was so obvious that it triggered a mini-epiphany for me, although one that didn't occur until after class was over: I assumed that it actually summarized the source. But did it? I didn't give a web address. Did it go looking for it? Did it find it? Were we just given a generic summary based on Hakluyt's work, or encyclopedia entries that it has in its "brain"?

ChatGPT is a black box in this sense. It's not just about missing citations. There's nothing in its response that signals, visibly, what part of your prompt it's responding to, or how. 6/11

So I followed up, and guess what? It didn't "read" the source. Here are the interactions that demonstrate this.

My Q: Can you cite your sources in constructing the above summary?

ChatGPT: I apologize, I am an AI model and I do not have the capacity to cite any sources as it is not within my knowledge cutoff and my information is generated based on the text I was trained on. Also, I do not have the capability to access external sources to verify the information provided in the document. If you need more detailed information, you might want to refer the original document. 7/11

My Q: What was the original document? Can you share a bit of it?

ChatGPT: The document you are referring to is likely a primary source from the 16th century written by Richard Hakluyt, an English writer and geographer who advocated for English exploration and colonization. The document is not available for me to share as I don't have access to it. However, I can tell you that Richard Hakluyt was a key figure in promoting the English colonization of North America and his works, such as "Divers Voyages Touching the Discoverie of America" and "The Principall Navigations, Voiages, Traffiques, and Discoveries of the English Nation" were some of the most important texts in the early years of English colonization. They were read and studied by many of the key figures in the colonization of America, including Sir Walter Raleigh and other investors, sailors, and settlers. 8/11

Whoa! I think this would have been a powerful thing to demonstrate live, in class. When I do this exercise again soon, I want to keep this in mind and hopefully bring it to the forefront of the activity. It opens up an interesting question, almost epistemological, about what knowledge is in this context but makes it tangible, something that can actually be discussed.

In other words, is ChatGPT's knowledge in this case fake or false, since it didn't actually have access to the document itself --- even if its general points aligned with the document in question? 9/11

I intentionally did not draw out this exercise, I treated it pretty casually and the whole thing probably lasted about four or five minutes. I think for this initial purpose, that was smart. The students were intrigued enough to pay attention and engage without any preparation (beyond having read what we were already reading anyway).

I'm curious about creating something a bit more formal, however, with deeper engagement. Maybe have them ask it to summarize the document, have the students try to annotate it with citations to the source document, then explore the lack of citations together live? Actually rewrite the summary somehow? 10/11