Ethics Announces AI Policy
Ethics: An International Journal of Social, Political, and Legal Philosophy has published its policy regarding the use of artificial intelligence (AI) by authors, editors, and reviewers.
Douglas Portmore (Notre Dame), editor-in-chief of Ethics, had shared some preliminary thoughts on the matter in an earlier issue of the journal, discussed at length here.
The official policy is divided into guidelines for contributors and guidelines for editors and reviewers, and notes that “editors reserve the right to sanction those who fail to [comply with the policy], possibly by banning them from reviewing for or submitting to Ethics.”
Here are the guidelines for contributors:
AI tools cannot author or even co-author a submission. Authors and co-authors must take full moral and legal responsibility for their submissions; they must take responsibility for the assertions that they make, the arguments that they proffer, and the sources that they cite or fail to cite. Since AI tools cannot take responsibility, Ethics will not consider submissions that have been authored or co-authored by an AI.
Authors must never take credit for work that’s not their own. Taking sentences of text from a generative AI tool and presenting them as your own words is plagiarism—at least, insofar as we take plagiarism to be a form of intellectual dishonesty in which one takes credit for work that’s not one’s own. Using generative AI to come up with a list of objections to a thesis or argument and presenting them as your own is also plagiarism. Consequently, authors who use text, images, or other content generated by an AI in their submission must be transparent about this, disclosing which tools were used and how. In cases where an original human source cannot be identified, authors should, then, include something like the following note: “I first became aware of this objection through the use of ChatGPT, OpenAI, April 16, 2025, https://chat.openai.com/chat.”
It’s important not only that authors avoid taking credit for work that is not their own but also that they give credit where credit is due. The problem with merely citing an AI tool—say, as the source of an example or objection—is that the AI tool may not be the original source. The original source may instead be an author whose work was used to train that AI tool. Thus, attributing some example or objection to ChatGPT could be just as problematic as attributing an objection to your colleague when all they did was tell you about some objection that Rawls raised. Thus, authors may need to track down the original source of an example or objection generated by an AI tool and cite it.
The editors of Ethics value human creativity; we, therefore, value work that presents the author’s own original ideas and insights more than work that presents ideas and insights that don’t originate with the author. For this reason, content that does not originate with the author may be deemed less desirable and publishable—at least, other things being equal.
Here are the guidelines for editors and reviewers:
Submissions to Ethics are confidential. Thus, editors and reviewers must be careful to maintain confidentiality and protect the author’s intellectual property. Just as sharing a submission with a colleague would violate confidentiality, so too would uploading it (or any part of it) to external services that lack the access controls needed to preserve confidentiality. The editors of Ethics have agreed not to upload submissions to such external servers nor to use AI tools to assess them.
Those who agree to review for Ethics do so under the implicit understanding that they will provide the editors with their own independent assessment of the submission, an opinion that is based solely on their own reading of the manuscript. Thus, it would be inappropriate for a reviewer to base their assessment of a submission, even in part, on an AI’s summary of it. And it would certainly be inappropriate to use an AI tool to generate any part of their report or assessment.
Discussion welcome, as are links to or excerpts from other journals’ AI policies.
Say I have written an awkward, run-on sentence and I am having trouble editing it down, so I give it to ChatGPT and ask it for help, and it gives me a list of alternate wordings. I select the best one and edit my paper accordingly, perhaps making adjustments to the AI-generated text here and there. Would this use of AI be prohibited under the new policy?
Say a referee writes a very negative report but then worries that their tone is too dismissive and might be demoralizing to the author, who is likely to be a grad student. So, they ask Claude for suggestions on how to soften their language and, after carefully assessing each of its suggestions, decide to implement some of them. Given how common unnecessarily harsh referee reports are, this would seem to be a good practice. Yet, Ethics policy appears to forbid it.
So you take this to be an instance of using an AI tool to generate your report (or assessment). I don’t. I take this to be an instance of using an AI to edit a report that you generated for tone.
I presumed that this sentence applied to both contributors and reviewers: “Taking sentences of text from a generative AI tool and presenting them as your own words is plagiarism”. Does this only apply to contributors? Or, is it your position that using a generative AI tool to rewrite a sentence to better capture the tone you want not “taking sentences of text from a generative AI tool”?
Right. I don’t think that this is an instance of “taking sentences of text from a generative AI tool.” Rather, it seems to be an instance of using an AI tool to edit text that you generated. In any case, I encourage you and others to read the entire policy, including the introduction that Justin left out. The introduction indicates that what we’re trying to do is to lay out some principles that we think that people should follow: (1) don’t take credit where credit is not due, (2) given credit where credit is due, (3) protect confidentiality, and (4) give us your own assessments and base those assessments on your own reading of the text and not on a summary generated by an AI tool. If you keep in mind that these are the principles that we’re looking for people to follow and interpret the rest charitably, I think that you’ll find that the policy is quite moderate and sensible.
After reading your clarifications in this thread, and re-reading the text of the policy, I can see that the policy isn’t intended to be as opposed to all uses of AI as it had seemed on my first reading. The initial wording sounded to me like it was asserting that copying and pasting a sentence from the output of an LLM was ipso facto plagiarism, but now I see that “plagiarism” is intended as a substantive characterization of *some* such copy-pastes, but not ones where (say) I had put in a roughly-written 12 sentence paragraph to have it edit it down to a tightly-written 4 sentence paragraph.
I’m still not totally sure how to interpret “Consequently, authors who use text, images, or other content generated by an AI in their submission must be transparent about this, disclosing which tools were used and how”. I have a recent paper, co-authored with Michael Nielsen (https://link.springer.com/article/10.1007/s10992-025-09814-6) that includes several diagrams (on pages 1109, 1110, 1112, 1120, 1122, 1123, 1129, 1130, 1132). For all of those diagrams, I gave Claude a verbal description of the diagram I wanted and asked it to write up the TikZ code for the diagram. In most of those cases I either just directly copy-pasted that code into the document and saw that it gave what I wanted, or did a slight tweak of the labeling, though for the diagram on p. 1110 there was an extended back-and-forth where I had it create various simplified version of this so that I could figure out the syntax and then used that understanding to modify the code into the version that I wanted.
I can’t quite tell if the policy would ask that I include this clarification in an acknowledgment footnote. I don’t know if it counts as “taking credit for work that’s not my own” to include a diagram based on code that I never learned how to write, or if this is just like using the LaTeX system to typeset my paper even though I don’t know the details of how LaTeX works. (In previous years I wouldn’t have included diagrams as ambitious as these, since I didn’t know how to create them, but I absolutely have copy-pasted the code for tables in LaTeX found on MathExchange into some of my papers, and not thought of it as plagiarism, because it’s basically like writing up a truth-table.)
It wouldn’t bother me to be asked to include this sort of acknowledgment of AI assistance in drawing diagrams and editing text, but I can see how different people might read the text of the policy and some might think that this kind of AI assistance was banned, some think the policy just calls for acknowledgment of it, and some think the policy allows these sorts of uses to go unstated.
You write: “I can’t quite tell if the policy would ask that I include this clarification in an acknowledgment footnote.” The policy suggests that you need to be transparent about your use of AI. So, I would interpret the policy as asking that you include that clarifaction. You also write: “I don’t know if it counts as ‘taking credit for work that’s not my own’.” If you explain that you used an AI to help generate the diagram, then that wouldn’t be taking credit for work that’s not your own. If you, however, give your readers the impression that you came up with this diagram all on your own when you didn’t, then you would be taking credit for work that’s not your own.
A wonderfully good practice at the small, small price of a piece of your humanity and your own ability for expressing basic human empathy using your own capacities in the written word. What a joy to be a humanist and man of letters in this new millenium.
Yes, it would be wonderful if reviewer 2 were “a humanist and man of letters” with a refined “ability for expressing basic human empathy”. Sadly, that is not our world.
My view is that AI tools should be explicitly acknowledged when used to re-word a sentence. I’m not sure whether this practice should be permitted if explicitly acknowledged.
Different words mean different things. Different orderings of the same words can mean different things. When AI tools are used to re-word a sentence, they can alter the meaning – even if only slightly – of a sentence. For this reason, it’s difficult to restrict the use of AI tools to mere presentational (as opposed to substantive) support, and authors should explicitly acknowledge AI tools when used in this way.
It would be prohibited under this policy only if it were right to interpret this use of ChatGPT as a form of intellectual dishonesty in which one is trying to take credit for work that’s not one’s own. Personally, I don’t think this is the correct interpretation. Using a dictionary, thesaurus, usage guide, copy-editor, spellchecker, grammarchecker, or ChatGPT to help with editing one’s writing (even without acknowledgement) doesn’t seem to be a form of intellectual dishonesty in which one is trying to take credit for work that’s not one’s own. The work that you’re taking credit for consists in the ideas that you are conveying in your own words. I take it that the words are still your own in this case.
As a non-native speaker, I really appreciate this clarification, and I totally agree with Portmore’s interpretation. I have the sense that when I use Chat GPT to reword or correct my sentence, that sentence is still fully my own–because it’s me who decides whether and to what extent to adopt the suggestions on the basis of my sense of English. (I even improve my English through this practice, somtimes) This is so just as the resulting sentences are my own when I use dictionaries, thesauruses, etc.
Not sure. But this sounds like an excellent opportunity to hone your writing skills, struggling through to a solution you uncover yourself.
Writing a paper presents many such opportunities to hone many skills, whether it’s one’s typesetting skills of deciding where to include line breaks and which words to hyphenate at the end of lines, or honing one’s design skills by coming up with a bespoke format for the bibliography, or honing one’s spelling skills by checking your own spelling of every word by hand. We’ve decided that some of these skills, despite being valuable, aren’t so valuable that we want to force people to develop these skills while they’re writing a paper that is meant to do something else. We’ve decided that it’s better to let computers catch all our spelling mistakes, and use standardized formats to organize the bibliography, rather than making humans spend their attention on these (even if humans might come up with different bibliographic systems that are more effective for particular papers).
If someone thinks that the point of writing a philosophy paper is to come up with the ideal wording for the surface presentation of an idea, they might object to using AI for this. But if someone else thinks that the precise wording is primarily important as a tool for conveying some further meaning, they might think it’s all to the good to have AI tools that provide lots of alternative wordings that a human can quickly choose between, rather than making the human come up with the wording themself every time (and sometimes end up with a suboptimal one because they didn’t consider enough alternatives).
I’m a fan of this policy (thank you, Ethics!), but I wonder about the paragraph regarding referee confidentiality, where it says that reviewers violate confidentiality if they upload a submission “to external services that lack the access controls needed to preserve confidentiality“. What counts as sufficient “access controls”? For example, in the CSU’s current contract with OpenAI, the materials submitted by faculty (and students) to their CSU ChatGPT accounts are not shared with OpenAI, may not be used for LLM training, and so on. That seems like sufficient access control to maintain confidentiality, no?
Of course, these are the same folks who simply stole work product from us and other authors in training their LLMs, so maybe we shouldn’t trust the contract with the CSU. (I don’t.) But if we take the contract at face value, is confidentiality not necessarily an applicable concern?
See, for example, “Is My Data Safe?” here: https://www.csun.edu/it/software-services/chatgpt. (I guess we should keep ignoring that “data” is plural.)
“That seems like sufficient access control to maintain confidentiality, no?” Perhaps. At least, for now. Should in the future we find that tech people at CSU still have access to what faculty are uploading and are selling some of that stuff on the black market, then no. As we mentioned in the preamble (which Justin didn’t include above), we have no hope of anticipating how things will develop. The point then is to provide principles. One principle is that you need to protect the author’s intellectual property. If you can do that with the way that you’re using these services, then great. If not, then you should rethink things. The responsibility for protecting the intellectual property of the author of a submission that you’ve been asked to review is yours. So don’t ask me if the controls at your institution are sufficient to do that; you need to make sure that they’re sufficient before you upload the author’s submission to these services.
Cyber security professional who was originally trained in part in philosophy here.
“Preserve confidentiality” is not very helpful to say: in some places (e.g., the Canadian federal public service) one has to first do a data classification exercise (in our case, done by executives responsible for the data) and then a cyber security professional like me (using that and other matters) can determine a likely way to protect it from confidentiality problems (i.e., what control instances are needed based on some patterns and some professional judgement). I would also add that the currently hyped systems will (with moral certainty) destroy data *integrity* too, and this may be even more important – and harder to deal with – than the confidentiality. As it happens, only some of these controls are related to access, broadly speaking.
I would also point out that the supply chain security information is riddled with reports of the big companies lying / misinforming / getting caught out otherwise about their supposed terms of service.
Impressive. Very nice. Now let’s see Ethics‘ internal procedure for correctly identifying AI use in its submissions.
I read this as having a pretty anti-AI slant. By contrast, I’m more inclined to think “so long as you cite it, it’s probably fine.”
I don’t really agree with the idea that ChatGPT was trained on something, so by citing to ChatGPT, we’re effectively endorsing ChatGPT’s plagiarism of others. That’s not exactly what’s going on. For two reasons, I guess:
I also don’t really agree with it from the reviewer side. Say some journal sends me an article, and I wonder whether the article is appropriately situated within the literature. I haven’t read *all* of the literature, so maybe I’m missing something. I could start trying to read everything, or I could say “ChatGPT: here’s an article I’m trying to understand, can you help me figure out if it’s missing any major views?” And it’s like “Yup, Jones (2024) said something similar, but the author doesn’t seem aware of that paper.” Now what am I supposed to do? Did I “cheat” by using ChatGPT to tell me about Jones (2024)?
I guess my inclination is to write, in my report, that the new paper should have talked about the older one and didn’t (ChatGPT, 2025). It feels like barring me from using that research tool weakens my ability to write an effective review, or at least to verify whether I’m missing anything important. It also feels to me like the *author* would benefit from the “enhanced” review as well.
Finally, what if I say “ChatGPT: I’m working on a new paper on X. Please tell me the ten most important papers on this topic, and give me summaries of all of them.” And then I decide which of those to invest it. Is that allowable under this policy?
Or what if you’re an editor and want to know whether a paper was plagiarized. Wouldn’t ChatGPT be one tool (among others) that you’d want at your disposal?
Again, cite to it literally every time you use it. But taking powerful research tools out of the hands of researchers doesn’t seem like the right approach, to me.
Thanks for the conversation!
Dear Fritz,
I think that you’re being very uncharitable.
Yes: “*Lots* of ideas came from somewhere else.” But you should not give the reader the impression that an idea came from you when it didn’t. Do you disagree?
Where does the policy suggest that using an AI to find the source of an objection and then presenting that objection while citing the original source of that objection is plagiarism?
Where does the policy suggest that a reviewer would be cheating if they used an AI tool to find out if a submission is overlooking any important literature? The principle that we’re concerned that reviewers abide by is not “read all the literature yourself” but “protect the author’s intellectual property. So don’t upload a confidential submission to an external service unless you’re confident that they have the access controls needed to preserve confidentiality.
“What if I say “ChatGPT: I’m working on a new paper on X. Please tell me the ten most important papers on this topic, and give me summaries of all of them.” And then I decide which of those to invest it. Is that allowable under this policy?” Why wouldn’t it be? Where’s the intellectual dishonesty? Where’s the attempt to take credit where credit isn’t due?
Sorry. The third to last sentece should have been: Why WOULD it be?
When you ask ChatGPT where its ideas came from and it provides sources, is it accurate? I would have thought that for many things you can get ChatGPT to say, it cannot give an accurate account of its sources (and indeed nobody can, because the process of going from sources to output involves steps that obfuscate the link between some particular source and some particular output).
It’s answer, like everything else it produces, is a guess at what an answer to such a question would look like. It is not “telling the truth” about its reasoning because it has no criterion for truth beyond the statistical likelihood that its answer will be found acceptable.
If a team of medical researchers uses generative AI to develop a new vaccine, no one will complain as long as the researchers disclose how the vaccine was created. I believe some of the problems discussed in Ethics are critical to humanity and urgently need to be solved. Suppose the only way to do so is to collaborate with an AI tool – the AI helps the philosopher check the validity of an argument that is so complex that no human could grasp all the details. The philosopher lists the AI as a co-author — the AI contributed to the paper by solving a logic problem.
As I understand the new policy, this paper cannot be published in Ethics, correct? Or is it just the generation of plain English that counts as “authorship”? That is not how it works in other disciplines.
That’s right: an AI tool cannot take the sort of legal and moral responsibility that’s required for authorship. But why list the AI as a co-author? It was a tool that the author used to check the validity of an argument. It didn’t author anything.
In other disciplines (medicine, physics), people (but not AIs) are routinely listed as “authors” without having contributed any text. Many journals require a special section explaining who did what: “X collected data”, “Y wrote the text”, etc. Perhaps philosophy journals should be open to this, too?
>an AI tool cannot take the sort of legal and moral responsibility that’s required for authorship
Do you not allow posthumous publication? Publication by people who have cognitively declined to a drastic state since writing?
Your imagining “an argument that is so complex that no human could grasp all the details” isn’t a counterexample. It’s hard to imagine a paper in Ethics that turns on a proof so large that no human could check it, but suppose there were such a submission: LLM-based chatbots simply aren’t good enough at identifying valid arguments to merit trusting them in such a case. A theorem-proving or proof-checking algorithm is a different matter, but the policy doesn’t prohibit formal philosophy papers that use algorithms or simulations as part of the research.
Something that is related to international relations that depends on the (1976!) computer assisted proof of the four-colour theorem? 🙂
Martin puts his finger on something that deserves wide consideration. Some people seem to object to the use of AI in philosophy on the grounds that it dehumanizes it: philosophizing is a core human endeavor that requires years of training, insight, and effort. Using an AI, arguably, has none of this.
But I think that this devalues philosophy in the other direction. The questions that philosophy is concerned with are, as Martin aptly puts it, “critical to humanity and urgently need to be solved”, and if we don’t use all the tools available to us to answer these questions, we aren’t recognizing the value to humanity of its answers. By emphasizing process over result, this treats philosophy as to some extent a game – though, admittedly, a very rich and at times rewarding one.
“the AI helps the philosopher check the validity of an argument that is so complex that no human could grasp all the details”
It is precisely in situations where there is no possibility of human confirmation that an LLM cannot be useful at all.
Finally. Clear, ethical (moral) human reasoning gone back to.
But, save your applause – it had to be reasoned out, which does take time, and has.
And with those garnering un-ethical wealth from their scrappings and plagiarism still having monied and un-ethical backing, and much to lose: may yet be at whisper in collusions, for the keeping of their theft, w/ one another.
We hope & pray mankind will yet come out atop.
No AI policy can make everyone happy, but this is perhaps the most sensible one I’ve seen for a journal that doesn’t want AI-written papers. Nice work, Doug!
For those who don’t mind reading AI-written papers—or having AI read it for you, more likely—it’s already been suggested that you could start your own journal for that. If so, I think there’s a good case for requiring AI to be listed as the author or co-author, contra most policies: https://ethics.calpoly.edu/AIauthors.htm
Three quick thoughts on this.
Ironically, this is probably an AI generated cartoon.
“I kind of think using Claude or Gemini etc to turn bullet points into prose just a rather extreme form of using it as an editor, which is apparently ok?”
I think you touch on an interesting point here, the fact that it is difficult draw a precise line between writing and editing. However it seems to me that turning bullet points into prose would almost necessarily need to involve enough new writing that it probably crosses the line into writing. Though i suppose there would be exceptions where the bullet points contain a significant amount of well developed prose.
On the other hand there are things that we probably generally consider editing that could potentially also cross the line into writing according to my conception above. For example, someone here suggested cutting a 12 sentence thought down to 4 sentences. It seems to me that this would be a murky gray area as it would depend on how exactly the cuts are made. I can certainly imagine a scenario where a cut of that magnitude could add some information in subtle ways or is reworded to the extent that it really doesn’t make sense to call it the authors words.
Yeah, I agree.
I think it’s worth distinguishing two questions in these cases of, let’s call it extensive editing.
The more I think about these cases, the more I think we’ve been systematically under-acknowledging editing work for a long time. And these cases of extensive editing are probably ones where (a) LLM help should be acknowledged, but also (b) human help should be acknowledged and traditionally has not been.
Perhaps more journals (including the one I edit) should be requiring acknowledgments of editorial assistance. Unpaid editorial assistance is usually not flagged, but hopefully included in the list of “Thanks to…”. Paid editorial assistance is usually passed over in silence. Maybe both practices are wrong.
Yes, I was appalled to learn that it’s quite common not to acknowledge, e.g., the graduate student RA who compiles an index, etc.
There should be alt-text in the graphic. Not sure if that gets lost on this kind of platform?
Also, the graphic isn’t mine but found online. As it looks to be AI-generated, I don’t know how (or who) to credit that.
Isn’t there a glaring contradiction between the first and last sentences of the first paragraph?
“AI tools cannot author or even co-author a submission.”
“Since AI tools cannot take responsibility, Ethics will not consider submissions that have been authored or co-authored by an AI.”
The set of submissions excluded by this policy should be empty according to the first sentence.
“Cannot” in the first sentence refers not to impossibility but to impermissibility (regarding submissions to the journal), to my mind.
I would expect Portmore of all people to know the difference between cannot and ought not. But maybe it is actually confusingly written rather than contradictory.
I think it means something like explicitly now they cannot be designated as authors in the publication Ethics
That would make a lot more sense but then why not phrase it like you do?
No idea. I read it your way first and like you thought it contained that blatant contradiction. The whole thing is very…oddly unclearly written, still more so for being an explicit policy. On my first readthrough, I interpreted it to be basically banning all AI usage in any part of the writing process; then came to the comments section to find the editor clarifying that he meant nothing of the sort, and the policy is really all about taking a stance against people getting credit for ideas that they didn’t think up by themselves
I don’t understand how you and others can legitimately interpret the policy as “basically banning all AI usage in any part of the writing process” The policy explicitly says that “authors who use text, images, or other content generated by an AI in their submission must be transparent about this.” How does one go from the idea that we require you to be transparent about your use of AI to the idea that we are prohibiting its use?
Sorry about this. What I should have said is that AI tools cannot count as author or even co-author in the relevant legal sense. That is, they can’t count as author or co-author for the purposes of signing the relevant legal documents that all authors and co-authors must sign before we can publish their paper. What’s more, they can’t count as author or co-author for the purposes of taking more responsibility for, say, failing to cite a relevant source. But, of course, AI tools can be the author (that is, the originator or co-originator) of some paper. You can just ask the AI tool to write a paper on given topic.
Thanks for clarifying. That makes sense!
Overall this seems like a pretty reasonable policy. I was struck by this paragraph, however:
“The editors of Ethics value human creativity; we, therefore, value work that presents the author’s own original ideas and insights more than work that presents ideas and insights that don’t originate with the author. For this reason, content that does not originate with the author may be deemed less desirable and publishable—at least, other things being equal.”
This suggests that, other things being equal, the editors of Ethics prefer to publish work that doesn’t contain significant expository content or quotations of other authors. Is this the intended policy? For that matter, I would argue that most of the ideas and insights in a given paper don’t originate in the author of that paper. Philosophy is a multi-millenial conversation and ideas emerge in that conversation in ways that make it difficult (if not impossible) to determine from whom they originated.
I can’t speak for all the editors of Ethics, but I value the human enterprise in which humans, in dialogue with other humans and following certain norms of scholarship, come to better understand the world and their place in it through an exchange of ideas with each other, ideas that they themselves come up with. Thus, I care about both the source of the ideas and the process by which the understanding is achieved. I value human creativity. And I value the interpersonal interaction in which one person sincerely takes a position and presents reasons for it, and then another person applies the principle of charity to understand this other person’s thoughts. The two, then, engage in a sincere effort to understand each other and to critically engage with each other’s ideas. And the hope is that they will, through this process, come to better understand the world and our place in it. For similar reasons, we might object to what Rima Basu has called “bullshit philosophy.” Of course, other people may care only about the result, and not about the source or the process. I just think that it’s reasonable to care about these things. Also, I think that it’s important that the most interesting ideas presented in a paper belong to the author. I think, then, that many of us would prefer an article in which the main insightful points were those of the author rather than reports of other people’s insights. Imagine, for instance, a paper in which the most interesting bits are the parts that are filled in for the ellipses in schemas such as the following: “In conversation, philosopher X has suggested….” If the most interesting bits are due to philosopher X as opposed to the author, then let’s publish philosopher X’s paper, not the author’s paper, which merely reports on the insights of others. In any case, a more direct answer to the specific question is: Yes. As noted in our information for contributors, Ethics isn’t big on publishing expository stuff unless the exposition has “significant implications for ongoing theoretical work.”
I think this has been asked already, but how (if at all) will you screen for AI content? I appreciate the policy, don’t get me wrong. I am just curious how this policy will be implemented given the lack of reliable/accurate AI detection tools. If there is no way to reliably detect AI-generated content, there is no way to act on this policy.
Ethics is the best journal in ethics. People who submit there have the highest ethical standards. They will simply abide by the policy. Unless they’re cynical act-utilitarians who think they can make the world better by sneaking a marginally better AI-written paper undetected.
I don’t think the second sentence follows from the first. Furthermore, although the first sentence may be true, the second probably isn’t. Or was the whole thing intended ironically?
Obviously a joke.
The policy provides guidelines that we expect people to follow. We have no explicit mechanism for uncovering undisclosed and problematic uses of AI. But we also have no explicit mechanism for uncovering other forms of academic dishonesty. For instance, we have no mechanism for uncovering whether an author has, say, plagiarized their students’ ideas. As we note in the document, “we articulate the principles with which all Ethics editors, reviewers, and contributors are expected to comply. The responsibility for complying with them lies with the individual.” It’s not our job as editors to make sure that authors don’t use AI in problematic ways any more than it’s our job to make sure that authors don’t steal the ideas of their students. It is our job as editors, we think, to try to provide some guidance on what principles our authors and contributors should be following when it come to this new and potentially problematic technology, which is what we’re trying to do with this policy.
Thank you.
The way I see it is like those signs in city parks some places that read that the park is closed to use at such and such a time at night. These are there to allow the police to do something if something untoward happens during the “dangerous time”.
So, does this mean it’s okay to submit a translation of my own writing using Google Translate, Grammarly, ChatGPT, etc.? If I use ChatGPT to translate my writing into English and submit it, is it okay to not disclose that I used it?
I hope that I have done enough to clarify things. Please keep in mind that each editor has his or her own individual views. This policy is an imperfect compromise of our differing views. In any case, I have to move on now. So I won’t be monitoring this feed anymore. Thanks for the discussion.
According to Harold Lasswell “Politics is the exercise of authority, influence and power to decide who gets what and when.” This, for those on the margins, there are those in power who decide who will have access to tools associated with Artificial intelligence…
Having just taught an upper-division AI ethics class, I’ve been able to glean something about what matters most to students – the human-to-human connection in philosophical exchanges. They love the movie “Her” because it really gets at what happens to human beings when they, say, have someone (or something) else write their personal cards and letters for them. Whatever else AI does for them, they’re most worried about losing connection with other actual people.
Douglas Portmore captures this for me in what he wrote, below: “I value the human enterprise in which humans, in dialogue with other humans and following certain norms of scholarship, come to better understand the world and their place in it through an exchange of ideas with each other, ideas that they themselves come up with. Thus, I care about both the source of the ideas and the process by which the understanding is achieved. I value human creativity. And I value the interpersonal interaction in which one person sincerely takes a position and presents reasons for it, and then another person applies the principle of charity to understand this other person’s thoughts. The two, then, engage in a sincere effort to understand each other and to critically engage with each other’s ideas. And the hope is that they will, through this process, come to better understand the world and our place in it.”
I don’t know whether this settles a lot of the great issues raised in other comments, but it seems like a moral anchor of some kind.
In conversations with students, that is definitely the sense I got. Going by their stated views, most of them seem very wary of AI in general (not just Gen AI). But revealed preferences are a thing. The behavior of (at least intro level) students tells a different story.
This raises interesting questions. I accept some, perhaps all of the arguments here, but I also doubt that this policy could be enforced in any meaningful way (certainly not without making many false accusations). As an individual, it might make sense for me to impose such an ethical standard on myself. But how can it make sense for an institution to impose a policy that it has no way of implementing?
As Portmore has observed, the same is true of policies against stealing students’ or colleagues’ work, fabricated data, etc. As with plagiarism, it doesn’t give you the ability to do anything in advance; but it does give you grounds for retraction if the wrong facts eventually come to light.