Philadelphia, PA

Dear Conroy & readers,

No. The use of a chatbot needs to be acknowledged.

H.G. Callaway

ChatGPT is a powerful tool that can be used to generate creative text formats of text content, like poems, code, scripts, musical pieces, email, letters, etc. However, it is important to remember that ChatGPT is not a human being. It does not have the same level of understanding and knowledge as a human being, and its output is not always accurate or reliable.

If you are using ChatGPT to generate answers to questions, it is important to be transparent about this. You should disclose to the person you are communicating with that you are using AI assistance, and you should carefully review the output of ChatGPT to ensure that it is accurate and appropriate.

I can see that using AI assistance for completing assignments is like using a calculator or a dictionary, but if the aim of the assignment is to think through some ideas and present an argument, the use of AI assistance can undermine the purpose of the assignment. This is also true of assignments about doing long division yourself or finding synonyms. It is certainly unethical to reproduce work that is not yours as your own without acknowledging the source, as someone else may be lead to believe that you skills that you do not have. Of course, there is nothing wrong with using AI assistance to support research in the same way as you would use a search engine to suggest sources and summarise the literature. But critical appraisal of the results are always needed as the results may be incorrect or incomplete. If the objective is human learning, effort needs to be put in to examine and think through the source material for the assignment.

To me, if you use ChatGPT to produce answers to questions and then pretend you have written them yourself, you should be stripped of your PhD, UHD and/or professor titles.

The question I'd like to know is why someone would do this. At least the ones that I see called out are obvious (and easily detected by chat bot checkers) so make the person look as if they are unable to answer themselves, and often do not recognize the errors that are, at least these days, still common in many chatbot engines. This makes the person appear as if they are trying to mislead the questioner. So, why do they do this?

Daniel Wright – most people on RG answer questions because science is a collaborative communal project. The ones who copy and paste ChatGPT are not interested in being helpful, but in being apparently knowledgeable. They are the ones who used to write twaddle like "good question, very interesting" just so they could put it in their CV that they had answered 12,000 questions on RG. At least, that's my take. Why otherwise would anyone "answer" 20 to 30 questions a day, in a wide range of areas that they cannot possibly be an expert in?

But it's crunch time for RG. Either they step up to the plate with a strong policy on this or the whole credibility of the site goes down the tube.

Interesting, too, that these AI engines are also prone to "hallucinations", inventing facts and references that don't exist. See https://retractionwatch.com/2023/09/01/withdrawn-ai-written-preprint-on-millipedes-resurfaces-causing-alarm/

What has struck me about the references supplied to statistics answers on RG is their hopeless vagueness – one answer just referenced some basic books on SPSS without any indication of why or where they were linked to the answer content. But the case above demonstrates a far more disturbing sort of problem, where statements are supported by nonexistent references or references that have nothing to do with the assertion.

Of course, writing a doctoral thesis using artificial intelligence is against all ethics. But the question that is no less important is to what extent this technology can be used to prepare such academic works.

Is this post a puzzle to figure out which answer was created by AI ?

Sal Mangiafico, the post of my prime suspect was flagged as "AI Content" by CopyLeaks.com, and the entire post was marked red. Ironically, that post stressed the importance of transparency when posting AI-generated content. Go figure.

Bruce Weaver, I tried two that looked AI-ish to me, and both came up AI in copyleaks.

This feels like, in the U.S. --- for certain statutes related to specific crimes --- where the prosecutor has to establish mens rea : evidence that the accused has knowledge that what they are doing is a crime.

Sal Mangiafico – the mens rea point is very relevant. I've just exchanged messages with a developing country biostats person who posted an AI-generated answer. She didn't see anything wrong with this – the AI text explained more clearly and with better organisation than she would have been able to, since English is not her first language. I would put her down as nescient – not knowing something that a person in her position would probably not know – rather than ignorant – not knowing something that a person in her position ought to know.

Academic teaching on plagiarism, personation (assuming the identity of someone you are not with the intent to deceive) and AI is now clearly behind the rapid advances in the field.

This might be of interest to those following this thread: https://www.nature.com/articles/d41586-023-02477-w

The papers identified appear people who don't proofread the output of their LLM generated text. I think several of the people who answer on RG do not proofread the LLM answers as evident that many of them have errors. I think one issue not covered in this thread is that by posting an answer that, by default is assumed to be by the poster, you are also liable for the errors in a public and archived place. Given the current fantasy-prone (and confident) responding of LLMs this may come back to affect people if they don't declare that this was an LLM generated response. One of the things that worries me is that the next step after individual academic's publish or perish dilemma, is institutions turning a blind eye to these issues in order for them to move up the league tables.

Oluwaseyi Ayorinde Mohammed , first, reading about your experience was useful for me, thank you for posting. Second, I am curious about the role of the adviser, who presumably was reading drafts throughout the dissertation period and if this was an empirical dissertation discussing these. Maybe the role of adviser is meant just to read the final product. I also thought, what if the adviser did think something was odd, and their reasons to have you read it rather than confront their student with the results from an AI checker. I am sure that there are lots of other cases like this that do not have someone as diligent as you to question the student.

Oluwaseyi Ayorinde Mohammed , I don't think anyone's saying you can't use AI tools.

Just like you can use books and websites.

The question is about using the results verbatim --- or nearly verbatim, or even partially re-written --- with the implication that the poster is the author, without acknowledging the source of the text.

I find that having this discussion about the behavior of academics with positions at universities to be baffling. My question is why are people posting AI results without acknowledgement. And then when they are called out, they refuse to add acknowledgement to their posts.

I think, no.

Theodore John Drizis – please tell us your reasoning. Join the discussion.

I did just send an email to ResearchGate. I know people have tried before. And I know I've made these points before in these forums.

But it seems that if RG doesn't institute some kind of policy allowing flagging of AI content or a policy statement about AI content that will also allow reporting the content, that the forums will just degrade to a worthless collection of Chat-GPT answers.

In terms of getting useful, relevant answers, the structure of the Question and Answer forums on RG already put it at a disadvantage to some of its rivals. (No separation among topic forums, no moderators, no downvoting, sometimes no core of experts correcting responses).

Here is a gobsmacking reason why we should not trust AI. This paper, written at least partly by AI, was retracted because – wait for it – two thirds of the references were invented. Not just the citations. The AI wrote paragraphs and paragraphs summarising research and opinions that were simply not real.

Have a look at the paper here. The paragraphs of bogus material are highlighted.

https://spectrumnews-web-assets.s3.amazonaws.com/wp-content/uploads/2023/10/19124325/Litvak-Private-Equity-Investment-Friend-or-Foe-highlighted.pdf

Interesting. ChatGPT is revolutionizing the world of thinking, and then it is not. A friendly scientific adversary of mine e-mailed me a summary of what ChatGPT thought of his theory - in an attempt to convince me he's right and I am wrong. [I will not elaborate but the 'heat' in our discussion is on Wheeler's 'charge-without-charge' idea: he believes charge can be further reduced to 'geometry', and so I do not.] The 'ChatGPT summary' of his theory is friendly and flows nicely but, when reading it, I noticed ChatGPT's AI glossed over rather fundamental questions and failed to make historical and scientific associations that I make. I pointed this out to my friend, together with some remarks from Johann Hari's recent book (https://www.amazon.com.be/-/en/Johann-Hari/dp/0593138538/ref=asc_df_0593138538/): how our human mind works - its 'thought' processes - remains largely unknown to us - especially the non-linear, non-analytical and subconscious aspects of it. Thirty years ago - when I was young and did not need much sleep - I would not have said that, but, as an old man struggling with tough scientific questions (look at this thread around a seemingly easy question: https://www.researchgate.net/post/Can_matter_be_created_out_of_light) - I've experienced how refreshing good sleep can be. Just as an example, I thought for weeks about how to motivate my E=4ℏω relation when trying to develop a model of the proton. Then I abandoned and, bang, the next morning all is clear. I am intrigued by Ronán Michael Conroy's suggestions AI tools just go wild. I have not looked into it, but it does not surprise me. Not at all. AI may or may not be a threat. At this point of its development, it looks merely 'entertaining'.

Jean Louis Van Belle, are you familiar with Guy Claxton's book "Hare Brain, Tortoise Mind"? Judging by some things you said, I think you might enjoy it. (John Cleese certainly did, FWIW. See the 2nd link below.)

• https://www.harperacademic.com/book/9780062032119/hare-brain-tortoise-mind/
• https://www.edutopia.org/hare-brain-tortoise-mind

Cheers,

Bruce

Dr. Conroy,

It strikes me as a small part of a much larger issue with principled inquiry in the academy at large. Whether one presents AI text as their own, or lifts paragraphs wholesale - to fabricating study results or merely giving themselves 'favorable' interpretations (leading to failure to replicate that may or may not be caught) - it is all merely fraud.

Too often we are fooled by our own sophistication; falling prey to our own expertise. In some rather horrifyingly common instances, the one who commit the fraud does so with full knowledge, but reacts with such deep apathy so as to raise the question of some level of sociopathy - these are so lost in sophisticated rationalizations that even the suggestion that they have done something wrong strikes them as confusing.

The answer to the question is an obvious, abundantly clear "No.", but the existence of the question exposes much the same as the 'replication crisis' or more recent heavily sensationalized plagiarisms at the most prestigious levels of Western academia.

It is universal, in my mind, that the 'ethical acceptability' of an act is inversely proportional to the effort required to 'explain' the ethical acceptability of that act - one rarely needs to contrive a justification for what is just, or to rationalize what is rational.

I am nearing the day when I will be able to say, without compunction or reservation, that academia is bereft of principle; that day has not yet come, but I believe I can see it approaching.

The method of fraud is irrelevant; fraud is fraud.

Spencer.

In my opinion, it is not ethical, except if it shows the source, as reference.