Dikarya permits the use of artificial intelligence tools in preparing manuscripts, but AI must assist authors rather than replace the intellectual work of authorship.
Our policy is based on four principles:
Authors must stand behind every idea and every sentence
The authors are responsible for everything submitted under their names, regardless of whether AI was involved in producing or editing it. Authors should understand, verify, and be willing to defend every statement, interpretation, citation, and conclusion in the manuscript.
AI cannot be an author and cannot take responsibility for scientific claims.
Writing is part of thinking
Scientific writing is not merely the transcription of ideas that already exist. The process of deciding what to include, organizing evidence, finding the right words, noticing inconsistencies, and revising an argument is itself part of scientific reasoning.
Authors should not use AI to circumvent that process.
Using AI to improve grammar, suggest clearer wording, identify awkward passages, or help organize material can be useful. Asking AI to do the substantive thinking and then accepting the resulting manuscript with little intellectual engagement is not.
Authors should put more effort into a manuscript than they ask readers to put into it
A scientific paper represents a request for the time and attention of editors, reviewers, and readers. Authors should respect that time.
AI makes it possible to generate large amounts of plausible prose almost instantly. The ability to produce text cheaply does not make that text worth reading. Authors should carefully edit, verify, shorten, and improve AI-assisted material before asking anyone else to read it.
As a general principle, the effort spent preparing and refining a document should exceed the effort expected of someone reading it.
Longer is not better
AI has a tendency to turn simple ideas into unnecessarily long documents. Dikarya values clarity and information density over length.
If a short instruction or a few sentences contain essentially all of the intellectual content, expanding them into several pages with AI does not make the work more substantial.
A useful test is:
If a document was generated from a short prompt and adds little beyond what was already contained in that prompt, consider sharing the concise version instead.
Appropriate uses of AI
AI tools may be used for tasks such as improving grammar and readability, suggesting alternative wording, helping reorganize text, checking consistency, assisting with formatting, or helping authors explore how to communicate ideas more clearly.
They may also be used more extensively during drafting, provided the authors subsequently perform the intellectual work necessary to evaluate, verify, revise, and take responsibility for the resulting text.
AI output should never be treated as an authoritative source. References, taxonomic information, factual statements, calculations, sequence interpretations, and other scientific claims must be checked against appropriate primary sources or data.
Disclosure
Authors should disclose substantial use of generative AI when it materially contributed to the preparation of the manuscript. Routine spelling, grammar, formatting, or minor wording assistance does not require disclosure.
The purpose of disclosure is transparency, not punishment. Use of AI by itself is not grounds for rejecting a manuscript. The relevant questions are whether the work is accurate, original, appropriately sourced, intellectually substantive, and genuinely the work for which the listed authors are willing to take responsibility.
The standard
Dikarya does not attempt to determine authorship by detecting whether particular sentences "sound like AI."
Instead, we ask a simpler question:
Did the authors use AI as a tool for doing their work, or as a substitute for doing their work?
The former is welcome. The latter is not.