Recommendations on the use of artificial intelligence tools in fulfilling study requirements
Introduction
This document presents a set of voluntary recommendations for the use of artificial intelligence ("AI") tools in fulfilling study obligations and requirements at Masaryk University. It does not address the use of AI tools in research, which is another broad domain for the application of these tools. The recommendations build on the Statement on the Application of Artificial Intelligence in Teaching at Masaryk University, which in some points they develop and supplement with specific advice and suggestions. Given the dynamic developments in the field of AI, it is expected that these recommendations will be continuously modified, refined, and supplemented. The recommendations concern in particular AI tools used in text preparation, but analogy can also be made with image and/or sound processing tools. The intention is also to provide scope for the application of the specificities of disciplines and study programmes when citing and referring to sources of information.
In general, AI-based applications[1] should be used primarily as an auxiliary tool in fulfilling study obligations. They can serve well, for example, to check grammar and stylistics, to suggest the structure of the text, but they can also be used as a tool for stylistic reformulation of the text, to improve readability and clarity, to formulate summaries of longer passages of the text, as an aid in the preparation of research or to search for sources of published information, etc. AI tools, such as LLM-based chatbots ("Large Language Models"), are not recommended to be used to directly formulate the text itself, or to generate conclusions, arguments, findings or claims contained in the text, which should be formulated by the author independently. They are also not intended to replace the information available in the original sources, but they can help with searching. The completion of a written assignment is to be based primarily on data and information that the learner seeks from original sources, where the source used or its author can be clearly referenced. It must be stressed that the author of the work is solely responsible for the way in which AI tools are used and referenced.
In particular, when using AI tools, it is recommended to follow the rules set by the institution at which the assignment is prepared. If the institution does not have such rules, it is recommended to consult the person responsible for the referencing (e.g. qualification thesis with the supervisor, seminar paper with the lecturer, etc.). In the case of qualifying theses, it is recommended to add a sentence to the usual statement of origin:
“I declare that I have used AI tools in accordance with the principles of academic integrity and that I refer to the use of these tools appropriately in my work.”
Other similar wording may be used according to the industry or the supervisor's recommendation.
Lecturers are encouraged to provide space for learners to actively use AI-based applications in the preparation of learning tasks and the fulfillment of study obligations where appropriate. The process of completing and presenting the assignment is always an essential element of the assignment. It is through this that learners acquire relevant knowledge and skills. Assessment of learners' performance should therefore be based primarily on the process of producing the required output, and only then on the output itself in terms of form and content. It is also recommended that greater emphasis be placed on learners' presentation of the assignment and on the discussion of the content of the assignment, where learners need to demonstrate the skills and abilities to independently summarise, explain and defend their processes and conclusions.
Examples of AI tools that can be used in written work and how to reference their use
The Statement calls for transparency in the use of AI tools in written work and recommends openly declaring or citing AI tools in qualification texts.
The generally accepted criteria for (co)authorship include not only a significant creative contribution to the creation of the work as such, but also the approval of the final version of the final form of the work and acceptance of (co)responsibility for the work as a whole. Since AI tools do not create the final work in this sense and cannot knowingly approve or take responsibility for it, they cannot be listed as co-authors. The author in the context of copyright law is currently only the natural person who created the work. The work must be the author's own intellectual creation, reflecting the personality of the author,[2] who was able to express his creative abilities through decisions made on the basis of his creative freedom[3] and was able to put his "personal touch" on the work when it was created. This can happen at any of the various creative stages of the realisation of a work, which are conception, execution, and editing.[4] Although AI tools such as generative AI cannot be "authored" as such in this respect, in the context of the ethical level of academic work and academic integrity, their use needs to be declared and referenced in certain cases.
Elements of AI are now part of a wide range of tools and applications that can be used in learning tasks such as word processing. Therefore, AI tools can be used in different ways in the field of study task processing. Depending on the type of AI tools used, the appropriate referencing and citation method should be chosen. In general, wherever the use of an AI application directly affects the content of an assignment, such as a written paper, the use of the application must be declared. If the AI application is only used for formal editing of the text, it is not necessary to declare the use of such a tool.
Among the applications using AI used in the processing of study tasks, especially written work, are:
- 1) Applications for spelling, style and grammar checking, predictive keyboards, etc. (including e.g. Grammarly). It is not necessary to declare the use of these applications.
- 2) Applications used to create links, organize source documents (professional publications, articles), provide an overview of existing publications, etc. (e.g. Google Scholar, etc.). It is not necessary to declare the use of these applications separately.
- 3) Machine translation tools (e.g. DeepL, Google Translate). It is advisable to refer to the use of this tool if the translated text is part of a written work. The author can do this, for example, by referring to the relevant passage of the text where the translation obtained in this way has been quoted or paraphrased. In addition to the usual citation of the source of the text, it is appropriate to supplement the citation with a further reference to the use of the relevant translator. The author should give a reference to the source of the original text cited according to the citation standard and add a note, which may take the following form: “Translated by [APP NAME/TOOL].”
- 4) Generative AI tools such as text generators, information extraction applications, automatic text summarization, chatbots, applications integrated into web search engines to summarize web content (SciSpace Copilot), applications to summarize and search scientific resources (Consensus), language models based on LLM technologies (e.g. ChatGPT, Bing AI, Google Bard, Perplexity, Copilot, etc.). In the case of the use of these tools, referencing depends on the way in which the textual output thus generated is used in the work. This may be the case, for example:
- a) The application of generative AI provides background information (e.g. summaries, etc.) and suggestions for the structure and content of the text, etc., that are not used in the written work in a literal text or paraphrase. In this case, it is appropriate to provide a statement in the text of the thesis, such as:[5]
“In the preparation of this work, the author used [APP NAME; APP WEB ADDRESS] for the purpose of [REASON]. After using this tool, the author has reviewed the content and takes full responsibility for it.” - b) The application of generative AI provides cues to the content of the text, which are used in the written work in verbatim, direct quotation, or paraphrase. In such a case, it is appropriate to provide the relevant passage with a citation, which here is essentially a declaration of the use of the relevant AI tool. The citations are made in accordance with the citation guidelines of the institution at which the assignment is produced. If the citation guidelines do not contain instructions for citing AI tools or their outputs, another citation standard is followed by analogy, currently e.g. APA,[6] The Chicago Manual of Style,[7] MLA[8]. We recommend following updates to the relevant citation standards for citing AI tools. A relevant prompt (query text for the application of generative AI) and/or summary can be included as part of the reference.
- a) The application of generative AI provides background information (e.g. summaries, etc.) and suggestions for the structure and content of the text, etc., that are not used in the written work in a literal text or paraphrase. In this case, it is appropriate to provide a statement in the text of the thesis, such as:[5]
Upon consultation with the supervisor of the written work, depending on the type of application and the way it is used, and last but not least based on the practices in the respective discipline or study programme, referencing the use of a generative AI tool (ChatGPT, etc.) can be solved according to the following recommendations:
- general declaration (see 4.a above);
- direct reference according to the selected citation standard (see 4.b above);
- a description of the use of generative AI in the appendix and/or in the methodology section of the thesis. The author will prepare a summary or summary of the ways in which the relevant tools of generative AI have been used in the written work. He/she may also provide a commentary on the limitations that exist in the use of AI tools in the context of the relevant written work and how he/she has dealt with them. In particular, he/she will provide information on the application used (with references), a summary of the relevant prompts (either verbatim or in summary form), the outputs of the generative AI application (again, either verbatim or in summary form) and explain where and how he/she has applied the outputs so generated in the text of the written work;
- the author attaches to the written work a copy of all prompts and relevant outputs of the generative AI tool in the form of a dataset, which is then referenced in the output (written work) in an appropriate way.
The referencing procedures proposed here can be combined in various ways depending on the circumstances. The author of the output can also archive the prompt and output strings of the generative AI, especially for final papers (in the form of screenshots, copies of text, etc.).
Conclusion
There is no need to worry about the use of AI-based tools in solving learning tasks - they are already a common part of word processors, internet search engines, and many other applications. These tools are becoming a common part of academic work, study, and research. However, it is necessary that the use of these tools and applications is ethical, creative, beneficial, and safe, as stated in the Statement on the Application of Artificial Intelligence in Teaching at Masaryk University. Therefore, we recommend following the principle of transparency so that the assessor of the study obligation, the reader of the written work, etc. can best understand how the work was created or how the specific study task was performed.
The group for AI in teaching at Masaryk University, in the responsibility of Mgr. Michal Bulant, Ph.D., Vice-Rector for Education and Quality (May 2023)
Contact: Mgr. Jakub Havlíček, Ph.D.
Notes and references
[1] An online course on the principles of artificial intelligence can be recommended “Elements of ai”.
[2] See the ruling of the Court of Justice of the EU C-145/10 – Painer, art. 88.
[3] See the ruling of the Court of Justice of the EU C-145/10 – Painer, art. 89.
[4] See Hugenholtz, P. B., Quintais, J. P. Copyright and Artificial Creation: Does EU Copyright Law Protect AI-Assisted Output? IIC - International Review of Intellectual Property and Competition Law. 2021, No. 9, p. 1201.
[5] See recommendations of Elsevier.
[6] See APA and APA 7th.
[7] See The Chicago Manual of Style.
[8] See MLA.
Picture: MACKENZIE, Mike. Artificial Intelligence & AI & Machine Learning. 2018-08-16 [2023-06-20]. Available under license Creative Commons at: https://www.flickr.com/photos/152824664@N07/30212411048/