Responsible Use of AI in Academic Writing
Subject to prior agreement with the supervisor, AI tools may be used in research work. However, AI must never replace the author’s critical thinking. The author remains fully responsible for the final version of the text and is required to disclose the extent of AI use and to cite publications indicated by the creators of the tool.
Before using AI tools, authors should review the applicable regulations, terms of use, licensing conditions, and related documentation, paying particular attention to copyright compliance, data security, and the quality of the tool’s output.
Authors should confirm that the AI tool does not claim ownership of content produced with its assistance and will not use submitted data for training or to expand its own resources.
The use of AI tools must be ancillary, meaning that such use (in consultation with the supervisor or instructor) is permitted only insofar as it supports the achievement of learning outcomes and does not serve as a substitute for the author’s work.
Agreements regarding the scope of AI use and the instructor’s consent should be documented, at minimum in an email correspondence summarizing the arrangements.
If the course learning outcomes do not include assessment of language accuracy, and the author has used AI for proofreading, they should disclose the specific tool used and the extent of its use.
The author remains fully responsible for the final version of the work, as AI tools can only support, not replace, the author’s contribution.
During the preparation of this work, the author used a tool (e.g., ChatGPT) to refine the language (e.g., stylistic accuracy). After using this tool, the author has reviewed and edited the text and assumes full responsibility for its content.
If the course learning outcomes do not include assessment of translation skills, the author may, after consultation with the supervisor, use machine translation tools. However, the author must always indicate the scope of such use (e.g., abstract, citations) and independently verify the translation, including completeness, language and stylistic accuracy appropriate to academic writing, and terminological consistency throughout the work.
During the preparation of this work, the author used a tool (e.g., ChatGPT) to translate selected parts of the text (e.g., abstract, citations). After using this tool, the author reviewed and edited the translations and assumes full responsibility for their completeness, linguistic and stylistic accuracy, and terminological consistency throughout the work.
The genre of the abstract in academic writing requires conscious and well-considered decision-making during the writing process. Automatically generated abstracts may fail to adequately reflect the full scope of the research conducted by the author.
Below are three example descriptions of AI tool use. It is recommended that a description of the tools used and the extent of their use be included as an appendix to the work.
Whisper (Radford et al. 2022)
A speech recognition and transcription tool based on language models, used for the initial transcription of collected data. The transcripts were subsequently manually verified and corrected in accordance with established conventions (cf. fdsaf). This use is consistent with standard practices in comparable corpus-based research.
Stanza (Qi et al. 2020)
A language-model-based tool for automatic text annotation, used for tokenization, lemmatization, and part-of-speech tagging of identified lexical units. This use is consistent with standard practices in comparable corpus-based research.
ChatGPT (OpenAI 2023, version 4o)
A large-scale generative language model used to generate, refine, and optimize Python code for the development of research tools. The generated code was verified, corrected, and extended. It was also checked for correctness and compliance with the adopted methods implemented within it, based on the relevant scholarly works cited in this study (see the section on tools). This use falls within the category of so-called AI-assisted (or AI-powered) programming.
