Use of AI tools to create content for academic papers

ChatGPT is just one of a new class of tools that can generate realistic natural language text and also program code upon a prompt by a user.

Consider this case: An author writes a paper and uses ChatGPT to generate parts of the related work section. The author does not mention this in the paper. Is this acceptable?

I would say it is plagiarism. The author plagiarizes from ChatGPT. I would argue that this is not acceptable. ChatGPT cannot be an author, but an author may also not plagiarize from ChatGPT.

Another case: The author writes a paper comparing code samples in different programming languages. The code samples are generated by ChatGPT. This would in my view be ok, since this is like displaying the output of an image processing program. The author would need to indicate with the code/text samples that they are generated with ChatGPT. The code/text samples are not part of the scientific argumentation in the paper. They are just objects of the argumentation.

Which standpoint should CEUR-WS take?

Comments are welcome.

2023-02-21: CEUR-WS published its rules on AI tools for content creation at https://ceur-ws.org/ACADEMIC-ETHICS.html

Some easy tests on whether a paper is related to computer science

The CEUR-WS Team has decided to insist that submissions need to be from the computer-science domain in order to be published with CEUR-WS:

https://ceurws.wordpress.com/2022/01/04/openaccess-ceurws-in-the-year2022-more-international-more-focused/

So, here are some easy tests for workshop organizers to see whether their event is likely producing a proceedings volume that is falling into the computer science category:

Q1: Are the authors of your papers affiliated to a computer science (CS) or information systems (IS) department?

Not every author have to come from a CS or IS department. But it should be the case for for overwhelming majority of papers, say 80%

Q2: Did your authors and reviewers previously publish in CS or IS outlets?

For CS, you can check whether they published in ACM or IEEE conference proceedings or journals or in well known CS series like Springer LNCS. For the IS field you can check whether they published in the journals listed in https://en.wikipedia.org/wiki/List_of_information_systems_journals. Another method is to check the so-called DBLP footprint of the authors/reviewers. Count the number of publications of the authors / reviewers in DBLP. Exclude papers that were published in CEUR-WS to avoid distortions. If the numbers are very low (less that 5 for authors, less than 15 for reviewers), then your event is probably not well-aligned with CS/IS.

Q3: Does the Call for Papers refer to typical CS/IS problems, i.e. related to the creation or evaluation of computing solutions (software, data models, process models, algorithms, …)?

If not, then you probably attract papers from non-CS/IS authors. Popular topics e.g. from nano-technology are not likely to attract papers from CS/IS authors. It may be that your event is about the crossing of CS/IS and another field, e.g. computational solutions for nano-technology problems. But then, the focus shall be on the computational aspects, not on nano-technology, physics, engineering, biology, medicine etc.

CEURART LaTeX template now produces PDF/A 1a

Starting from version 0.4.8, our LaTeX style CEURART produces PDF/A-1a. It used to produce PDF/A-3u, but since some national archives would only accept PDF/A-1a, we made this switch. The Word/LibreOffice templates are not affected. We advise authors to configure their Word/LibreOffice suite to also produce PDF/A-1a when creating papers for CEUR-WS.

The change is minor and should not have a great implication. To our knowledge, PDF/A-3u allows to include data tables in the PDF file. Since, no standards for the data formats are defined in PDF/A-3u, such inclusions are less suitable for archival.

If authors need to publish data sets with their papers, we recommend to either use open-access services that issue DOIs/URNs for such data sets and then link to the data sets from within the PDF file. Or authors can use the complex paper format

http://ceur-ws.org/Vol-XXX/paperX/

for publishing papers with small accompanying data sets/source code/models.

N.B. (2022-07-06): While the CEURART style now uses PDF/A-1a, images contained in an actual PDF and included packages and options can lead to a PDF file that is actually not compliant to PDF/A-1a.

Semantified CEURART HTML variant for open-access publishing

Since more than a year, we offer the CEURART style for papers to be published on CEUR-WS.org:

https://ceurws.wordpress.com/2020/03/31/ceurws-publishes-ceurart-paper-style/

This style is available for LaTeX (preferred), LibreOffice and MS Word.

However, a semantified (using common tags for meta data and semantic elements) HTML text may be more suitable for some papers, in particular from the semantic web community.

Would you volunteer to create such an HTML style? It should render similar to the 1-column version of the CEURART style and made available under a free and open-source license. Embedded scripts should be avoided.

Hope you find that a challenge! Let me know if you are interested!

Manfred

The poisonous effect of #paywall artefacts to #openaccess

Spoiler alert! This is more of a rant.

Recently, I was involved in the publication of an open-access paper that allowed to attach accompanying research artefacts such as data sets. The artefacts were all open-access (produced for that paper) but one artefact was not. It had been produced by the organizers of a workshop to specify a so-called “challenge”. The artefact was openly accessible but when the workshop gut published by a computer science society, it became copyrighted by this society and was moved behind its paywall.So, new open-access publications cannot just include this artefact in their accompanying data sets.

I regard this as a poisonous effect of paywalled artefacts. It creates insecurity with the authors about what they can publish. It slows down research. Do not accept such behavior of paywall publishers! Demand that your results are freely accessible under a generous license. We are researchers. We share!

#Openaccess: What is a truly international event for #CEURWS ?

As written in the previous blog entry from January 4, 2022, the CEUR-WS team has decided to re-focus more on workshops, in particular international workshops. So what does that mean? Let me give my personal view. Please use comment to express your standpoint! We can then hope to come to a common understanding about the direction of CEUR-WS.

So, a “truly international workshop” for me is a workshop where the participants and the program committee members come from a many different countries. One can argue what “many” means but you get the idea: the more the better. This avoids the creation of “bubbles” where local communities discuss scientific problems only with themselves. But I believe we need even more for a “truly international workshop”. I think that the participants (and PC members) should have a history to go to conferences/workshops in many different countries. They should get out of the “bubble” and expose their results to the scrutiny of an international audience.

The best way to achieve such a goal is probably to organize the workshop in collaboration with a well-known international conference. Of course, there are exceptions. We do not completely rule out conference proceedings. Our quality check may be a bit tougher for conferences but if the quality is very high, those proceedings are welcome! Likewise, workshops that are linked to a country and where most papers come from that country are not excluded. In such cases, the authors of papers should have a history to present their results on truly international events.

There has been a bit of an inflation of the term “international workshop/conference”. This dates back decades ago. The attribute “international” was meant to stress the importance of the event. I believe we should scale down the use of the word. If an event is local, then call it “national workshop” or just “workshop”. We shall anyway see from the composition of the accepted papers whether the submission is indeed international.

National events are great. They help young researchers to get experience. However, they should then also be published nationally. At least, the organizers should seriously consider this option. A publication with a national publisher is also accessible internationally. So, it is not lost. What may be more difficult is to get papers indexed in DBLP and SCOPUS. But that is only fair!

All researchers want to have their works visible as widely as possible. But then you need to put an extra effort to get your works published in highly visible outlets. And that is hard. It is poised to get harder on CEUR-WS.

Or what do you think?

#OPENACCESS #CEURWS in the #YEAR2022: More international, more focused

#CEURWS (CEUR-WS.org) will continue to provide free open-access publication of computer science proceedings. Based on the past experience, we want to further strengthen the international reach of proceedings volumes. This means for future submissions that we will check that the workshop (or conference) needs to be international in its target audience, the program committee, and the authors of the accepted papers. Some workshops may stem from national associations. We do not rule out these cases. But the call for papers shall target a worldwide audience of authors. This should also be reflected in a truly international composition of the program committee, to include well-known experts for the topics mentioned in the call for papers.

Another point of attention is the focus of CEUR-WS. We serve the computer science academic community. Sometimes, a workshop may have a particular cross-disciplinary character. Still, the papers should discuss the research questions from a computer science point of view. This is typically the case when authors have a computer science background.

CEUR-WS started as a publisher for workshop proceedings. Workshops have a rather small set of topics and they try to attract papers from the best experts for these few topics of interest. Later, CEUR-WS also started to publish some conference proceedings. We will continue to do so but the acceptance criteria for conference proceedings are more rigid. Conferences must be truly international and have a rigorous review process. We do not want to publish regional conferences (and for that matter regional workshops) anymore. (New) conferences with a very broad set of topics of interest should better publish elsewhere. There are plenty of alternative publishers!

The CEUR-WS Team has the authority to reject a submission if the pre-conditions are not met or the team believes that CEUR-WS is not the right venue for the submission. Since workshop/conference organizers have enough alternatives, we do not discuss such decisions. You can however make a rejection unlikely by following the rules and certain strategies to ensure the high quality:

1. Workshops co-located with well-known computer science conferences (IEEE, ACM, …). Well-known means that the conference is indexed in DBLP and has a long history.
2. The organizers of a workshop and the PC members are well-known in the fields represented by the workshop. They are mostly from computer science/information systems departments.
3. The program committee is international, as well as the group of authors

#CEURWS #OPENACCESS : Are we stuck with PDF forever? Is there a chance for semantically enriched papers?

Practically all papers published at CEUR-WS.org are published in the PDF format. This is now a quite old format, originally developed for printing. I still regard it as a good paper format, in particular when the PDF file is supporting navigation and easy lookup of references.

However, it is not a format that allows easy searching/querying of papers because it has no meaningful semantic annotation of its contents.

There are a few papers in CEUR-WS that were actually written in a semantically enriched HTML format. Will such a format become more popular for academic publishing? Or are we stuck with PDF for the foreseeable future?

I personally believe that papers should also be data, i.e. it should be possible to query the contents of a paper. There are initiatives like the open knowledge graph at TiB Hannover to represent the research questions, methods and results in a way to facilitate semantic queries across a large collection of papers. I find that very promising but apparently PDF does not support such machine-readable content at all.

#OPENACCESS #CEURWS adopts CEURART as mandatory paper style in 2022

Since mid 2020, we have been popularizing the CEURART style for paper to be published at CEUR-WS.org, see

https://ceurws.wordpress.com/2020/03/31/ceurws-publishes-ceurart-paper-style/

Submissions using other styles were still accepted in 2020 thru 2021, but editors got informed that they should move to CEURART for future editions of their workshop/conference.

This posting is a reminder to prospective editors to adopt CEURART for future submissions. Details and examples are available at

http://ceur-ws.org/Vol-XXX/

There are templates for LaTeX incl. Overleaf, LibreOffice and Word. The CEURART template includes the CEUR-WS logo. Consult http://ceur-ws.org/HOWTOSUBMIT.html#FAQ for our conditions to use the logo in published papers.

#openaccess: Understanding #CCBY as used by #CEURWS

CEUR-WS has adopted the Creative Commons CC-BY 4.0 license (https://creativecommons.org/licenses/by/4.0/) for papers and volumes published on CEUR-WS.org.

It is important to to understand the difference between the legal provisions of CC-BY (see above) and the academic rules for publishing papers. The latter cover among other plagiarism, which is not so much a concern of CC-BY. CC-BY is a very liberal license. It supports the open-access publication chain by authorizing the publishers and users to store, access and distribute the papers.

CC-BY does not prevent certain undesired uses. For example, a third person could select a number of CC-BY-licensed papers and republish them in a book, as long as the clauses of CC-BY are not violated. This may not even require consent by the authors. Clearly, this is against academic standards. So, sanctions should be considered within the rule system of academia, e.g. by designating such a publisher as predatory.

Overall, the advantages of CC-BY are stronger than its disadvantages. CC-BY is the defacto standard for open-access publishing. Let’s hope that we get an improved version CC-BY 5.0 where some of the concerns are addressed.