AI evidence extraction

Generative Artificial Intelligence and Responsible Authorship: Scientific, Ethical, and Legal Considerations

Artemis Chaleplioglou, Alexandros Koulouris, Eftichia Vraimaki · Preprints.org · 2025

AI-generated evidence extraction, verified across multiple analytical personas. Not a substitute for the peer-reviewed original.

9/10

Relevance

0/4

Quality (LMQS)

Evidence

Citations

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.20944/preprints202512.2507.v1

Methodology & findings

Study design

Cross-sectional bibliographic scoping review using multiple databases (Scopus and HeinOnline) with direct keyword search for publications 2022-2025, backward reference search from 2022-2025 results (covering 1970-present), and comparative analysis using VOSviewer network analysis.

Main result

The study found that "GenAI models cannot be held responsible for the content of scientific papers" and that "the responsibility and accountability for the content of any contribution lies with its authors, with particular emphasis on the corresponding author." Analysis revealed "a rapid increase in relative backward-referenced publications from 2017 to 2021" with an R² of 0.9919, indicating strong linear growth in AI-related scholarship. Additionally, "the influence of GenAI use is evident in scientific papers, as reflected by a rapid increase in the frequency of commonly overused GenAI-related terms," with characteristic words such as "delve," increasing "by more than 6 times."

Research paradigm

Critical interpretivism / Hermeneutic analysis

Author conclusions

The authors conclude that "GenAI is an invaluable and powerful tool for researchers; however, legal regulation, ethical compliance, and the protection of proprietary rights and content from unauthorized use are essential." They further state that "Human involvement in preparing papers should remain active, vigilant, and cautious when using GenAI systems, given the 'black box' nature of their responses and the risks of errors, misinterpretations, biases, or hallucinations." The comparison of scholarly and legal perspectives "reveals challenges related to traditional authorship, creativity, ownership, and copyright" which are "a key concern for all stakeholders involved in scientific publishing, including publishers, editors, reviewers, authors, and readers."

Risk of bias

Selection bias from backward reference search methodology (only includes cited works); Publication bias toward recent literature (2021-2025); Database selection bias (Scopus and HeinOnline only); Language bias (English-language publications primarily); Temporal bias in citation patterns that may not reflect current attitudes; Selection bias: papers published before 2022 without citations are excluded; Database bias: only Scopus and HeinOnline were searched; other databases not included; Citation bias: backward reference search only captures cited works, missing uncited research; Temporal bias: rapid changes in GenAI technology may render findings quickly obsolete; Language bias: likely emphasis on English-language publications; Citation bias: backward reference search only retrieves cited works, excluding uncited pre-2022 papers; Database selection bias: limited to Scopus and HeinOnline; other academic and legal databases not included; Temporal bias: direct search limited to 2022-2025; pre-2022 coverage only through backward references; Language bias: likely English-language publication bias; Publication bias: analysis based on published literature only

Limitations

The authors acknowledge that "two bibliographic databases were used in the literature review: Scopus for scholarly communications and HeinOnline for legal communications." Additionally, "a direct keyword search for reports published before 2022 was not conducted
instead, a backward reference search was used
This method retrieves works cited in papers published after 2022, all of which have been cited at least once
As a result, papers published before 2022 that have not received citations, and those without citations from this dataset, were not included." The authors also note that "ongoing developments in GenAI models, agentic AI, and multi-agent systems, along with improvements in LLM fine-tuning and the creation of GenAI-powered scientific tools, are rapidly transforming the environment of scientific computing, data analysis, and writing."

Open questions raised

The authors identify gaps regarding the need for reforms to guidelines for contributors, reviewers, and editors of academic journals to clarify authorship attribution and safeguard research integrity. They note the need for clearer definitions of authorship criteria in the context of GenAI use and call for stronger legal regulation and ethical compliance frameworks.
The authors identify that "This is the first cross-sectional study comparing scholarly and legal literature on the use of GenAI in scientific report authorship," suggesting a gap in comparative analysis between academic and legal perspectives. The paper also highlights ongoing uncertainty regarding GenAI's role in research design, data analysis, and experimental validation, as well as the need for clearer guidelines on authorship attribution and disclosure requirements.
Further examination of how legal frameworks across different jurisdictions (U.S., E.U., China) specifically address GenAI authorship liability
Development of standardized, comprehensive guidelines for GenAI disclosure across all academic publishers
Investigation of detection methods for GenAI-assisted content in scientific manuscripts
Analysis of the long-term implications of GenAI integration on research integrity and scientific validity

Extracted from: pdfAgreement 85%

Explore related topics

Responsible AI Use Research Integrity Literature Review