cross-posted from: https://piefed.world/c/tech/p/1311057/people-prefer-stories-written-by-ai-especially-when-told-theyre-written-by-a-human
People gave the highest ratings to stories generated by AI that they were told had been written by a human, and were unable to tell the difference between human- and AI-created stories, according to new research published by Cambridge University Press.
Is there a place where we can do that test? The article links to a download, but it tells you which is which.
“public assumptions about the capabilities of AI are increasingly out of date”
writing stories is another to add to the list of things AI does better than humans
I agree that the anti AI crowd is consistently out of date or misinformed about AI. It’s an echo chamber which makes them believe AI is less capable than it it and the anti AI crowd should work on that. More of the real arguments and less of the “AI bad at tasks”.
That’s like saying people prefer lots of highly processed food with heaps of sugar and no nutrients. It’s definitely true in some ways, but not a good thing to organise your life or society around
That’s a great analogy!
Why? What harms come from the AI generated stories to their readers?
Considering you immediately down voted my comment, I think it’s fair to ask: is that a genuinely curious question?
edit: fuck it, I’ve been thinking about this anyway, may as well write it down. A few points:
Firstly, a few decades ago a similar claim about junk food would have been very hard to find evidence for, because research on it’s detrimental effects didn’t gain much momentum until after it was obvious there was a problem. AI will certainly have a similar lag, so it’s unlikely that we’ll have hard evidence of it’s effects (especially in the wild) for years or decades.
That said, I have a few hypotheses for impacts on readers/reading:
- “quality” in this study is probably fairly highly correlated to “average” - text that reads simply and clearly, because it tries to be as close to cultural norms as possible is much easier to read. That removes barriers to reading, which to some degree makes it more fun to read. But it also makes it less challenging to read. Counterpose this with creative literature that often has a really strong author’s voice that can be a bit challenging initially, but beautiful it you persevere with it. I’m thinking like Ursula Le Guin, or Haruki Murakami.
- in the longer term, reading a lot of AI writing is likely to lead to a strengthening of the bias towards simple, generic language. In turn this could lead a reduced demand for really creative writing. By analogy, we’ve already seen something similar happen with hollywood films over the last couple of decades - more focus on generic (e.g. Marvel) and rehashed stories.
- AI use is already leading to more generic, less explorative science - see Hao, Q., Xu, F., Li, Y., & Evans, J. (2026). Artificial intelligence tools expand scientists’ impact but contract science’s focus. Nature, 1–7. https://doi.org/10.1038/s41586-025-09922-y
- Because LLM-produced writing inherently produces less diverse content (because it is an averager, with noise, see the Hao reference above), then it’s less likely to produce ideas that challenge the mind, and require the reader to engage critically.
- LLMs usage is already associated with reduced critical thinking capacity: Gerlich, M. (2025). AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking. Societies, 15(1), Article 1. https://doi.org/10.3390/soc15010006
- LLM’s well-known tendency towards “hallucinations” is all fun and games when it’s overt, but more insidiously, more subtle hallucinations happen constantly too, and go undetected. This can end up producing ideas/memes that are subltly wrong and potentially damaging, but get accepted by the reader because the sound pretty close to the truth (especially when you’re lulled into a false sense of security by a voice that sounds VERY MUCH like it knows what it’s talking about).
- Guardrails are garabage and easily bypassable and have already lead to deaths. While reading doesn’t have the rabbithole effect of using an LLM as a chatbot, it’s still capable of producing horrendous ideas and framing them as positive.
- I suspect that reading less diverse content also results in a degraded capacity for imagination in general, so it could lead to less good art in the world.
- Imagination is also necessary for solving major socio-environmental issues, and even for things like planning to prevent terrorism, so this could have broader ramifications in the long run. See Weick, K. E. (2005). Organizing and Failures of Imagination. International Public Management Journal, 8(3), 425–438. https://doi.org/10.1080/10967490500439883
I feel like a had more points, but that’ll do for now.
Plus obviously all the other GenAI complains - environmental destruction, copyright theft, EEE power grabs, associations with fascism in general…
- “quality” in this study is probably fairly highly correlated to “average” - text that reads simply and clearly, because it tries to be as close to cultural norms as possible is much easier to read. That removes barriers to reading, which to some degree makes it more fun to read. But it also makes it less challenging to read. Counterpose this with creative literature that often has a really strong author’s voice that can be a bit challenging initially, but beautiful it you persevere with it. I’m thinking like Ursula Le Guin, or Haruki Murakami.
The #1 thing I can think of is plagiarism.
AI stories are quite literally unoriginal works generated from original mediums and it’s harmful to all authors worldwide.
AI-giarism is still plagiarism.
https://lawreview.uchicago.edu/online-archive/plagiarism-copyright-and-ai
https://www.theatlantic.com/technology/2026/08/ai-companies-buying-used-books-for-data/688167/
https://www.sciencedirect.com/science/article/abs/pii/S1096751625000806
Did you read the stories?
“If you lie, people won’t notice”.
Quite the lesson to learn.
that’s a lot of psychology experiments right?
you make them do this and then watch that
you misinform them and see what they do with that information
The milgram experiment wouldn’t work if they told people that they’re not really killing someone or without “lying” they would have really needed to kill someone which is worse.
There’s also ethics involved like a debriefing after the experiment and the lying or deceptions can’t cause damage etc.
People are dumb and often do things that harm themselves and the world as a whole. That’s the reason AI is bad - not because we don’t like it, but because we shouldn’t like it, it makes us dumber, and is terrible for the planet on top of it all.
It’s like if I lit a bunch of tires on fire, bottled the smoke to smell as perfume, and justified it all by saying I found someone who liked the smell. It’s irrelevant, and doesn’t justify my actions at all.
As someone who reached early adulthood, before internet became mainstream, I have to say that everyone took a collective dive in the deep end of the stupid pool after social media was born.
AI will make us more stupid and intellectually lazy.
I’ve played around with perchance.org story generator enough to be legitimately impressed with it. I do think you need a human to direct the story and correct issues but it writes better than I do, it just isn’t as good at coming up with ideas and it frequently makes obvious mistakes.
Someday is now today.
