Research articles in the GPT era

As a journal, we are proud of Matt Duckham's vision of a community-led, open access journal for Spatial Information Science.That vision has some important cornerstones: that all papers are open access; that there are no charges for publication enabling anyone to publish in JOSIS regardless of their access to funds; and, that we remain community led, without the need to consider commercial interests.These principles come with some costs.All of the infrastructure behind JOSIS is run on a shoestring, and as editors we are not only responsible for shepherding papers through the review process, but also for the myriad hidden tasks involved in typesetting and publishing papers.One way of streamlining the process somewhat is moving from considering final accepted manuscripts in Word or Latex to only Latex file format.Although we are aware that there is a learning curve to using Latex, the availability of an Overleaf template makes this much less daunting than in the past.To make this transition easier, we are moving to consider initial submissions in any format, but will ask authors to agree to submit a Latex version should their papers be accepted.A second, and in many ways more existential challenge concerns the widespread adoption of large language models-or more informally, generative AI-such as ChatGPT and Perplexity, by increasing numbers of authors as writing tools.Many journals are now being confronted with the reality of scholarly written work involving the use of such technologies in their creation 2.More concerning than simple stylistic changes, such as word choice, is the potential for these models to generate factually inaccurate content, including citing nonexistent research literature.In addition, the risks of "accidental" plagiarism and copyright infringements are real.A more subtle issue is the notion of scholarly work as a unique, creative expression of researchers' thinking.The common saying that "writing is thinking" is, we think, highly pertinent to this debate.Our field is struggling to find reviewers for the deluge of articles being submitted to journals, and reviewers in turn have less and less time to review.Time savings, through the use of new tools to accelerate writing, will only exacerbate this problem.An important counter-argument is the leveling of the playing field for those whose first language is not English.If tools like ChatGPT really helped reduce inequalities in publishing, especially with respect to the underrepresented global south, then we would be on board too.But, at least for now, a much more likely path appears to be that even more papers are produced by already productive groups.For JOSIS, the potential benefits of publishing work that has involved the use of generative language tools in their writing, do not outweigh the risks at this stage.Therefore, in line with publications such as Science, we are instituting a policy that submissions must be written entirely by the authors without the help of AI writing tools, except for minor improvement of grammar 3.