AI evidence extraction

Generatiivinen tekoäly ja tekijyys tieteellisessä julkaisemisessa : kirjallisuuskatsaus nykytilasta

Iida Ahola · LUTPub (LUT University) · 2026

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

9/10

Relevance

1/4

Quality (LMQS)

Evidence

Citations

Methodology & findings

Study design

State-of-the-art literature review using systematic mapping study (SMS) methodology.

Sample

N = 135, 3 groups

Primary method

Qualitative systematic mapping study methodology using three-level relevance assessment (topic, abstract, full-text) with structured inclusion/exclusion criteria. No quantitative statistical analysis performed.

Main result

The study found that "generative AI tools cannot meet the criteria of gaining authorship due to their inability to assume ethical and legal accountability" and that "the role of scientific author is shifting from producing text toward supervising, evaluating, and validating AI-assisted outputs." The research demonstrates that while generative AI can enhance writing efficiency and accessibility, particularly for non-native English speakers, "human authors remain responsible for accuracy, originality, and integrity of the final work regardless of AI involvement."

Reports effect sizes.

Research paradigm

Interpretivist/Hermeneutic

Author conclusions

The authors conclude that "generative AI is not replacing scientific authorship but transforming it." They state that "The study suggests that traditional assumptions regarding authorship, originality, and academic integrity need to be reconsidered in this increasingly collaborative human-AI writing environment." Further, they note that "generative AI should not be treated as an invisible tool" and that "most scholars and policy frameworks support recognising AI as a contributory technology that should be disclosed transparently, while rejecting the idea of granting authorship status to AI systems themselves."

Risk of bias

Selection bias: Limited to peer-reviewed articles published 2023-2025 only; excludes grey literature and earlier foundational work; Database limitation: Search restricted to Springer Nature Link database; other major databases not included; Publication bias: Focus on peer-reviewed articles may exclude critical perspectives from preprints or non-traditional sources; Language bias: Implicit assumption of English-language publications given search string; Topic bias: Rapidly evolving field means included studies may reflect early/preliminary findings rather than mature evidence; Publication bias: only peer-reviewed articles included; emerging field may have limited publication coverage; Selection bias: database limited to Springer Nature Link; Temporal bias: rapid evolution of AI tools may render findings outdated quickly; Language bias: appears to be limited to English-language publications; Scope bias: focus on ethics may underrepresent technical or legal perspectives; Time-limited search (2023-2025) may miss earlier foundational work; Database limitation to Springer Nature Link may introduce publication bias; Peer-reviewed articles only may exclude grey literature and emerging perspectives; Rapidly evolving field means included studies may become outdated quickly

Limitations

"The thesis was conducted as a literature review and therefore relies on existing academic discussions rather than empirical observations." Furthermore, "because generative AI technologies and their integration into scientific writing are still relatively new developments, there is limited research focusing on their ethical implications
Much of the available literature emphasises guidelines, institutional policies, and technical concerns, while ethical discussions about authorship, accountability, originality, and intellectual contribution remain underdeveloped." Additionally, "as long-term empirical evidence on AI-assisted scientific writing is still limited, many discussions within the literature are more speculative than evidence-based."

Open questions raised

Future research should focus more extensively on the ethical dimensions of AI-assisted authorship and scientific writing. Future studies could explore how concepts such as originality, plagiarism, creativity, and intellectual ownership are evolving within hybrid human-AI writing environments. Additionally, more interdisciplinary research combining ethics, technology, education, law, and publishing studies would help develop clearer frameworks for responsible AI use in scientific communication.
Limited ethical research: "Much of the available literature emphasises guidelines, institutional policies, and technical concerns, while ethical discussions about authorship, accountability, originality, and intellectual contribution remain underdeveloped."
Need for empirical evidence: "As long-term empirical evidence on AI-assisted scientific writing is still limited, many discussions within the literature are more speculative than evidence-based."
Future research should focus on ethical dimensions of AI-assisted authorship, how concepts of originality, plagiarism, creativity, and intellectual ownership are evolving in hybrid human-AI environments, and interdisciplinary research combining ethics, technology, education, law, and publishing studies.
The authors identify several future research directions: (1) More extensive focus on ethical dimensions of AI-assisted authorship and scientific writing; (2) Exploration of how concepts such as originality, plagiarism, creativity, and intellectual ownership are evolving within hybrid human-AI writing environments; (3) More interdisciplinary research combining ethics, technology, education, law, and publishing studies to develop clearer frameworks for responsible AI use in scientific communication.

Extracted from: pdfAgreement 75%

Explore related topics

AI Manuscript Writing Academic Writing Literature Review