Recently I wrote a blog post on teaching writing to students of Spanish in the new age of AI: La Palabra Perfecta. It was written in Spanish, and specifically aimed at my own students, as well as anyone else who is teaching bilingual students to write in Spanish. In it, I muse over the distinction between human and artificial intelligence in writing, using as my inspiration a poem by Spanish poet Juan Ramón Jiménez. Here it is in English:
Intelligens1, give me
the exact name of things!
…Let my word be
the thing itself,
created newly by my soul.
So that all may go through me
those who do not know, things;
so that all may go through me
those who forget, things;
so that all may go through me
the very ones who love, things…
Intelligens, give me
the exact name, yours,
and theirs, and mine, of things!2
Juan Ramón Jiménez, Eternidades, 1918
I saw in Juan Ramón’s poem a powerful statement about human communication through the written word–the attempt to find the perfect words to express things. In my research, I found that others had turned to this poem in their own thinking about artificial intelligence in other contexts and their thoughts, along with the poem, helped shape my own: namely that AI is a tool we can’t avoid (and this will only be more true for our students in the future), but that we also need to support our students as they seek their own human voice in the written word, even when their words are not always perfect.
This idea came out of an initiative sponsored by the Provost’s Office on my campus to encourage faculty to tackle AI in our courses. Those of us in the initiative have been tasked with designing an assignment, or series of assignments, that introduces some AI tool in an appropriate context in our discipline. I’ve spent a little time playing with various generative AI chatbots trying a few small things. I have found them both amazing, and frustrating. One of the pressing questions for me has been: how can students ever learn to write, how can they discover their voice in writing, if an AI chatbot can write something seemingly perfect to them in a matter of seconds when it would take them hours, or even days, to produce something that would still be, in their minds, imperfect?
So, I wrote this post, in Spanish, for them, so that they could hear my voice, in Spanish (my second language). And then, after I published it, I gave it to three generative AI chatbots for feedback. I planned to share in class what the bots told me, and get student feedback on whether they agree. I will update this post after I do that.
But in the meantime, I am linking a copy of the responses to the prompt «I would like your opinion on this text. Please analyze and give suggestions for edits» (rather general, I know) from each of the 3 chatbots I consulted: Copilot, ChatGPT, and Claude.
Copilot and ChatGPT had some reasonable suggestions. Some of them I might have taken into consideration, in another circumstance. But they were all about form over substance. Copilot actually made a pretty significant error, correcting the spelling of «Intelijencia» that the poet spelled intentionally in this way. And that is where Claude, by Anthropic, got it really right. After reading its response, I thanked it, and said I had compared responses among three chatbots, and that I found that Claude was the most helpful. It asked me why, and here is the conversation that followed:
Claude: Thank you for the comparison feedback! I’m genuinely curious – what aspects of the analysis did you find most helpful compared to the other chatbots? Was it the balance of structural and content suggestions, or perhaps the way specific examples were highlighted? This kind of feedback helps me understand what makes analysis most useful for writers.
I noticed the text deals with AI and authenticity in writing, which makes your experiment with different AI chatbots particularly relevant to its themes. It’s interesting how you’re practically engaging with the very questions the text raises about AI’s role in writing and expression.
Me: The point about the spelling of intelijencia. That is how Juan Ramón spells it, although it is not standard. Not exactly an error, but your point about it being human, that was good
Claude: Ah yes, that detail about «intelijencia» is fascinating! It’s a perfect example of how Juan Ramón Jiménez’s intentional non-standard spelling actually enriches the poem’s meaning. It’s not a mistake but rather a conscious artistic choice that adds another layer to his exploration of language and authenticity.
This also connects beautifully to García-Máiquez’s point about the difference between machine and human «errors.» An AI would likely flag «intelijencia» as incorrect and suggest «inteligencia,» missing the poetic and philosophical intent behind the spelling choice. It’s similar to how e.e. cummings’ unconventional capitalization or Gloria Anzaldúa’s code-switching might be «corrected» by AI, when these choices are actually crucial elements of their artistic expression.
In your piece, this could be developed into a powerful point about how true linguistic creativity sometimes means breaking conventional rules – something your students might find encouraging as they develop their own voices in Spanish.
GLORIA ANZALDUA?? I am not sure I would have ever made that connection between Jiménez, e.e. cummings, and Anzaldúa, but it is spot on! I am not sure what to make of this. Is this a scary indication of how the computers will someday rule the world, or is it just a better way to brainstorm to help me make better connections? I’m not sure. So I’ll keep experimenting with AI, with my classes, and in my own work. Perhaps Claude can offer me some guidance?
