AI declaration field on the submission form OJS - future upgrades

Feature Requests
1

Describe the problem you would like to solve
Some of our editors are having problems with submitted manuscripts full of AI use including obvious hallucinations and fake references so we’ve been looking at different AI policies that journal editors might like to adopt. Many policies require authors to declare AI use on manuscript submission - including the full name of the tool, what is was used for, which tool, version…. This could include the use of assistive AI (editors’ choice).

Rather than authors declaring this in the submission documents, would it be possible to have an AI declaration box on the submission form? Editors could choose to display this metadata when published or use if for their own and for reviewer purposes.

Describe the solution you’d like
An option to include an AI declaration box on the submission form - including name of tool, version, how it was used/ reason for use. This could then be used as an AI use disclosure statement on the article page so the transparency of AI use is available to readers as well. An example of a Declaration of AI use https://doi.org/10.1016/j.jmrt.2023.08.246, https://doi.org/10.3390/ai6060128,

Who is asking for this feature?
Journal Administrator/ librarians providing support to editors.

Additional information
Editors may also require use of such tools elsewhere in the submission such as under methods but having a clear statement makes authors accountable from the start and makes the use of AI more transparent.

2

This is a very hot topic for many journals. It’s not particularly easy to modify the submission form, and different journals will ask for different fields. It’s probably easier to set a submission policy that governs how these tools are used, and require a statement within the article that answers specific questions. The ACM has such a policy here but I’m sure that many can now be found for other publishers.

3

Hi @Tuwpub,

Thank you for this feature request. Helping journals deal with AI-related submissions is a multi-facted issue. Our dev team might have some suggestions for how to best approach this in future iterations of the PKP software - I wasn’t able to glean if we have immediate plans for addressing such things @asmecher - do you know? :

The Committee on Publications and Ethics (COPE) has a few good guidance pieces on these issues:

https://publicationethics.org/guidance?query=AI&sort_by=relevance

1 Like
4

Hi @Tuwpub,

It sounds like you’re seeing AI being improperly used to produce bad science – and I wonder if requesting AI disclosures from authors who are using it that way would actually result in honest answers.

This is just my opinion, but the improper use of AI to generate apparently scientific content is not that different from authors using paper mills, and a solution could fall along similar lines – plagiarism checking, diligent peer review, and staying current with telltale signs to watch for. (I am not a journal editor and don’t have my finger firmly on the pulse for how this is typically handled.)

In terms of technical solutions, this will always be an arms race – but there are AI detection tools out there. For more than casual use they tend to be subscription-based. I’m not aware of any plugin-based integrations for OJS, but the Plagiarism plugin might provide a good starting point for anyone looking to develop one. It also looks like iThenticate (which is the service integrated via the Plagiarism plugin) is working AI detection into its products as well.

(Disclaimer: I’m skeptical about the value of generative AI used in the context you describe!)

Thanks,
Alec Smecher
Public Knowledge Project Team

5

Thank you, we’re just checking what is possible. We do have editors (particularly in STEM subjects) keen to allow Gen AI to be used in articles - with full disclosure and transparency - and I liked how the two examples I shared had declarations of use on the article page itself.

Including these statements within the article documents is the easy option which we will encourage editors to use for now.

Alec is right though when it comes to authors using AI when not permitted, they certainly wouldn’t declare it’s use! iThenticate is a little pricy for our editors so diligent manual checking is certainly being applied!

2 Likes
6

My two cents to this topic.

I may be wrong, but I don’t see much point in trying to detect the use of AI automatically, because as soon as a test is designed, the first thing new AI does is adapt to pass it… so it’s a never-ending race in which we will always lose.

Said that, we recently included this “AI policy” in most of our journals:

https://publicacions.uab.cat/en/journals/ai-policy

The policy is based on existing proposals (which aren’t many at the moment: WAME, publication ethics, etc.) and, in short, could be summarized as follows: “If you use AI: 1. don’t generate content without human supervision (and be aware of the AI issues), 2.sign a statement explaining EXACTLY what you have done with AI and how to reproduce it (models, prompts…), 3. never include AI as an author.”

A new item is added in our journal’s submission checklist so authors need to explicitly check to confirm they read and understood the rules.

To accept this AI-policy, we have added a new item to the journals’ submission checklist, so authors must explicitly indicate that they have read and understood the rules regarding AI.

Of course, as Alec said, this does not prevent cheating/understanding what they are signing… but, as with other issues (e.g. plagiarism…), you are giving your word, and in science, the author’s reputation is important, so we hope it will act as a deterrent and, well, also as a legal basis in case someone breaks explicit rules.

Cheers,
m.

PD: This post was originally written in Spanish, then translated into English using AI (deepL) and reviewed by a human whose English is far from native.

2 Likes
7

Hi @marc,

Thanks for sharing your experience with this. Could you clarify something:

You have two paragraphs here:

The policy is based on existing proposals (which aren’t many at the moment: WAME, publication ethics, etc.) and, in short, could be summarized as follows: ‘If you use AI: 1. don’t generate content without human suppervision (and be aware of the AI issues), 2. include a statement explaining EXACTLY what you have done, 3. do include the AI as an author.’

And:

The statement is based on existing proposals (which are not many at present: WAME, publication ethics, etc.) and, in summary, could be summarised as follows: “If you use AI, 1. be careful about the problems that arise, 2. sign a statement explaining EXACTLY what you have done with AI and how to reproduce it (models, prompts…), 3. never include AI as an author.”

Something may have gotten “lost in translation” here - can you clarify which one your journal uses?

My hope is that it is the latter, as the page reads;

1. Authorship of the article. The IA cannot be listed as author or co-author of an article. Although the legal status of an author differs from country to country, in Spain and in most jurisdictions, authorship is always attributed to a legal entity

That is a really solid policy, IMO, thanks for sharing - I’ve had a few folks ask me about policies about this, so I will be sending them this as an example.

-Roger

PKP Team

8

Ideed. :slight_smile: Thanks Roger!
And, in turn, my publication becomes a good example of an author who uses AI without due diligence or without any diligence at all. :stuck_out_tongue:
(In my defense, I would like to say that it is 38 degrees and I am not used to these temperatures).

That is a really solid policy, IMO, thanks for sharing - I’ve had a few folks ask me about policies about this, so I will be sending them this as an example.

Great! Thanks!

In fact, this AI policy is not applied in just one journal, but is recommended by the UAB editorial service (with 50 journals), and has been shared and adopted by our colleagues in the region… so feel free to share it if it is useful to anyone.

By the way, as the ‘version’ banner indicates, this is a ‘work in progress’ (in fact, given the speed at which AI is advancing, I don’t think any policy will ever be definitive), so comments and suggestions for improvement are welcome.

1 Like
9

I think that some overcomplication has crept in here - @Tuwpub hasn’t requested an AI-identification tool, nor a way to include a policy.

The request seems to be adding another metadata field on a submission form. Did the author use generative AI with this article; if yes, describe the generative AI used. Once an article is accepted, this metadata would display on the article’s metadata page, similar to this:

Screenshot 2025-09-01 110626
Screenshot 2025-09-01 110626724×176 17.5 KB

This seems no different to an author entering a title of an article, or specifying the section the article belongs to, another piece of metadata. If an editor can select whether this a required field or not, that’s a bonus.

3 Likes
10

Hi @mkmiec

Yes—that’s certainly part of it. But it’s easier said than done and dealt with, and this is an issue that has a broad number of implications that are worth discussing. Titles, abstracts, and other standard pieces of metadata have been parts of metadata schema for some time now, and I don’t think we’ve quite pinpointed how to best convey such emerging AI declarations as part of metadata. While it might be possible to include such information, it’s less apparent how such information would fit into standardized metadata schemas such as Dublin Core, that accompany the article even it is to move out beyond OJS (e.g. into indexing and abstracting services that, depending on how they are configured, may or may not be well enough equipped to handle such metadata). For example, do you list AI as a contributor (e.g. dc:contributor)? This is questionable from an intellectual property standpoint, as many jurisdictions acknowledge that AI/LLMs cannot be party to owning copyright. Or do you use a descriptive statement outlines its use (e.g. dc:description). We have to tread somewhat carefully here as embedding such statements in metadata in a non-standardized ways could have unintended downstream consequences, particularly for indexing and aggregation. The submission form within OJS is, by design, intended to be kept relatively simple, as a means of ensuring that unnecessary metadata does not make its way into downstream uses of the article metadata. Adherence to policies, which are also part of the submission process, is, in my opinion, part of this discussion but bit of a separate thing (but also intertwined) from the metadata itself - a number of these are internal checks and balances that the author needs to adhere to, but do often not end up as part of the public metadata. Suffice it to say, this is a multi-faceted issue, with many aspects to consider.

-Roger
PKP Team

2 Likes
11

I agree with @rcgillis that thinking about downstream impacts on the metadata is maybe the key thing here in terms of having a separate field (vs. just encouraging some kind of statement in the article itself). At the moment, there isn’t really an equivalent DC field in terms of giving that kind of detail about how the work was created. I am guessing dc.description of some kind would be the most appropriate but for now I think statements in the actual work make more sense than a separate field.

12

Thank you, excuse my ignorance - I’m a librarian not a developer, is it possible to have a custom field that doesn’t get exported via XML? Just a field that sits on the submission form and article page - such as in the examples shared.

If not - I totally understand that this new type of information doesn’t really fit current schema.

I ask because aside from AI statements, another type of metadata I’m interested in is local context licences https://localcontexts.org/ which recognises indigenous rights to their knowledge in published works.

2 Likes
13

Sorry I should have looked at your examples first – aren’t those just part of the article template? Like MDPI or Elsevier just requires that be part of the template like “Acknowledgements” or “Declaration of competing interest” – but they aren’t separate metadata fields…? FWIW The OpenAlex API button is a nice way to look at what search services/harvesters/etc are actually “seeing”, here is the record for one of the articles you shared above: https://api.openalex.org/works/W4386227153 – so there’s lots of metadata in there but the AI statement is just part of the full text of the article and doesn’t show up.

More generally since you’re interested I know I’ve seen some conversations around Local Contexts and PKP on GitHub (Add Traditional Knowledge Labels in OJS · Issue #6690 · pkp/pkp-lib · GitHub) but I’m not sure where that currently stands; given that the labels were developed in the context of museum collections some of the issues apply differently for journal article authorship in a way that I think folks are still figuring out but it’s definitely of interest to a lot of people! And rights information is something that’s already fairly standardized (e.g. canonical URLs and controlled names for CC licenses that we can put in dc.rights fields) so it’s likely that would be where the Local Contexts/TK labels integration might happen…? ie, I think everyone would want it implemented in a way that did get sent downstream.

I guess the thing to keep in mind is that the metadata fields contribute to description of the item in its journal context but also are key for its discovery and access downstream (e.g. in Google Scholar or Open Alex), and most people are going to be discovering the research somewhere other than the journal page itself – so it’s important to at least consider that side of things before adding specialized fields that won’t interoperate with other systems/standards.

tldr; as far as I can tell implementing something like what Elsevier or MDPI is doing above would be a change of template that the journal editors control, not actually a new OJS feature?

1 Like
14

Thanks Emily. It’s good to know there is potential to change the form to have the customised field. At the moment we’re on 3.3 and 3.4 and didn’t’ know about the editable form.

We’ve been looking at how to add Local Context labels to our D Space repository and it’s great to know that PKP has been looking into it for OJS. Colleagues have managed to add it to FigShare but it’s displayed in a tiny corner of the page.

15

Hmm, just to make sure we are on the same page – I don’t think I said anything about changing the form. The main info collected in the submission form is still going to be Title, Keywords, and abstract and then contributors – all of which are very standard metadata that is going to be exposed downstream. The examples you shared the AI statement seems to be part of the actual article template – like something that might be handled with the journal layout editing stage? Again from your example above, the author guidelines for that Elsevier article include very detailed instructions and template language for AI disclosure, but it’s not a separate field as far as I can tell: https://www.sciencedirect.com/journal/journal-of-materials-research-and-technology/publish/guide-for-authors#toc-8

1 Like
16

Thanks @ehopkins, for pointing out the TK labels issue in Github. That was a little while ago, and there were aspirations for developing a plugin. @kshuttleworth might know if there have been any movements on that in the SFU context. .Like was mentioned, I too have seen TK labels in cultural heritage contexts, but not really in publishing contexts, but I too am curious about how they might be used in more of a publishing setting. One thing I have noticed with TK labels is that they lack the machine-readable elements of other licenses, like Creative Commons licenses, which can be expressed more readily with metadata and transported with the digital item itself. Moreover, the TK labels are

Please note
if this conversation continues to relate to TK labels (I am just as guilty of taking it in this direction :wink:), we may need to parse it out into a separate thread, as to keep the thread focused on a singular issue, as per our guidelines.

I should note (and this is applicable to declarations of this nature - both AI and TK) that buried in that Github issue is also mention of a plugin: GitHub - tsv-fi/tsvVertu
that could possibly be repurposed for use for declaration fields. For use cases such as these, there is a desire for declarations or statements, but they are distinct from the standard metadata.

-Roger
PKP Team

1 Like
17

We had this same exact request from an editor.

I think the main requirement here is just in the submission form and how journals could collect this data form authors when they send the paper. How the declaration is shown or formatted is something that we can not handle as part of the publication metadata with no existing standards.

2 Likes
18

I honestly support the AI disclosure policy

19

Hi all! This is relevant:

STM: Recommendations for a Classification of AI Use in Academic Manuscript Preparation

This document presents a classification of various ways that AI can be used to assist with preparing academic manuscripts. It may serve as a framework for publishers to develop policies on how AI in manuscript preparation may be used and should be declared by authors.

I would love to see someone like NISO pick this up and make it an interoperable standard.

Regards,
Alec Smecher
Public Knowledge Project Team

2 Likes