Problems with AI responses on ResearchGate and elsewhere
Up until now, there has been an obvious problem with people 'responding' to a discussion or question by doing an internet search and then apparently cutting-and-pasting what they found, sometimes in large, perhaps illegal quantities, without reference and likely often without understanding the topic or their own such response. This will not likely advance actual understanding and may derail the discussion. It also adds to the unfortunate atmosphere of dishonesty. I have seen what appears to be far, far too much of this, as I am sure is the experience with so many of us. The internet, like a library, is a tool, not a substitute for thought.
Now we have AI. An AI response may further this problem: superficial, trivial, perhaps often verbose responses. It may increase the chance that the responder does not understand the topic, or even read his/her own response. We may as well have bots talking back-and-forth to each other. Superficial, trivial.
AI could help as a research tool, but it is a crutch which is liable to be used as a substitute for learning and communicating, and not an aid.
I suspect many are using AI responsibly and as a learning aid. But I suspect that too often, that is not the case.
What are your thoughts, in your own words?
I'm sure some may think it clever to use an AI response to this to see what can be detected. That just proves my point. You aren't doing anything yourself. AI cannot create, but only synthesize existing material. If people become too lazy to do anything themselves, future creativity dries up, and a tool that could have been helpful becomes a dead end.
Even if AI could eventually 'create' original thought in some manner, it will not be fully advancing human thought if humans aren't even really paying attention.
For now, we just have to worry about it being a lazy internet search. Superficial and trivial.
I suspect many are using AI responsibly and as a learning aid. But I suspect that too often, that is not the case.
What are your thoughts, in your own words?
This may be somewhat tangential to the discussion, but Richard A Williams recently posted links to his course policy on the use of AI plus an annotated bibliography to Statalist (see below). Anyone who is teaching courses nowadays may find them helpful. I'll certainly be sharing these documents with my students this year.
Cheers,
Bruce
- https://academicweb.nd.edu/~rwilliam/AIConcerns/AIPolicy.pdf
- https://academicweb.nd.edu/~rwilliam/AIConcerns/AIBibliography.pdf
- Dr. Williams' main page: https://academicweb.nd.edu/~rwilliam/
As you said: "If the internet is a library, then artificial intelligence is a book in that library." So, the issue is the person who enters the library. What is this person's goal? If it's to learn, they will feed their mind with the information they acquire from a book they choose from the library. If their goal is simply to appear "smart," then what good will a few good sentences copied from books do them? In that case, isn't the book more intelligent than the reader?
It will surprise no one that the topic of AI-generated replies (and what RG should do about them) has been discussed many times before. Here is one of those previous discussions (started by Ronán Michael Conroy in 2023):
- https://www.researchgate.net/post/What_should_ResearchGate_do_about_people_who_spam_questions_with_answers_generated_by_ChatGPT/1
Cheers,
Bruce
PS- Here is another discussion started in 2023:
- https://www.researchgate.net/post/Is_it_ethical_to_use_ChatGPT_to_produce_answers_to_questions_and_then_pretend_you_have_written_them_yourself/1
IYH Another approach is to go back to oral one-on-one student testing.
IMHO pedagogues have to foster AI use honesty. FWIW am partial to Daniel Miesler's AI Influence Level (AIL) declaration https://danielmiessler.com/blog/ai-influence-level-ail
To get accurate data about this sensitive topic where students are highly motivated to lie likely need to use the late Richard Hamming's "“How would you measure the amount of adultery per year on the Monterey Peninsula?” technique. From a large sample the slight excess of AI use above 1/2 gives the measure you want.
Daniyel -
I think you may be referring to the random response method, or something similar.
Yes, it would be nice to know what part of a work is AI and what part is human.
However, as when Bruce noted that student use of AI is a bit "tangential" here, I am a little more interested in the impact on discussion when someone injects a response which is not thought through because the responder may not really understand the topic and is just posting something they found without reading through it thoroughly themselves. This may derail the discussion. It has been a problem in the past, but AI has the potential to make this so much worse. Then, as I noted, "We may as well have bots talking back-and-forth to each other."
[In one thread that Bruce noted above, people were calling for the return of the down vote. No, please. That can get nasty, and wacky people down vote good responses.]
I encourage responders to read my thread introduction.
Thank you - Jim
Related to this problem is something I just came across again, which happens often: Someone will find a link to a related article, which would be good to attach and say a few relevant words about it and how it fits into the discussion. But instead, they copy huge extracts, and even though they make the source obvious, they are still likely breaking copyright laws and may even sometimes distract from the actual topic. In the latest case I found, it was very relevant, but verbose and I would think likely illegal.
In light of all this, I'd like to hear about your experiences regarding challenges to staying focused on topic.
What I do not understand (and complain about whenever I see it) is people posting AI-generated content without admitting the source. It would be one thing if someone thought they were being helpful, but quite another if they pretend that these were their own ideas.
• Gen AI: focuses on creating text, images, or other content when you ask.
• Agentic AI: goes further by making decisions and carrying out tasks to reach your goal.
- Badges
- Science topic
- Similar topics
- Social Science
- Media
- Journalism