Author:
Kait Krull

University of Tartu guidelines for using AI applications for teaching and studies

#studies

These guidelines outline the University of Tartu's general position on the use of generative artificial intelligence (AI) applications, including AI chatbots, in the learning process. The guidelines highlight the opportunities and risks of AI usage, provide instructions for the meaningful integration of AI applications into educational practices, and introduce best practices for using AI in thesis writing, including recommendations on how to cite AI usage.

The development of artificial intelligence (AI) has led to generative systems that can create text, images or other media so well that the result can be difficult to distinguish from human-generated content. As a result, over the past few years, generative AI applications based on large language models, including chatbots such as Microsoft Copilot and ChatGPT, have transformed how we think about text generation. This transformation has led to a debate about how to learn and teach at university, which skills are becoming obsolete, and which new skills are essential to avoid being left behind. 

A conversation with an AI chatbot begins when the user asks a question or makes a request as a prompt. Providing extra information and context in the initial prompt helps produce better results. The AI then generates a text response. The text responses may be enough to complete the conversation or raise a question or input for a continued dialogue.

Although the output of an AI application may initially seem meaningful and logical, it can contain errors. It may “cite” fictitious sources, make mistakes in logic, formatting, calculations, and grammar, or provide biased answers that fail to consider cultural differences or social norms, to name but a few. Additionally, the AI's output may not comply with data protection regulations and could contain inaccurate personal information. Therefore, all facts and references the AI chatbot provides should be carefully verified.

The responsibility for using the AI's output lies with the user, who must have the necessary knowledge to evaluate it. As long as we teach and assess the skills at the university, where AI applications (sometimes seemingly) appear faster and more effective than humans, we must also consider how to prevent academic fraud and integrate AI chatbots into teaching and learning in meaningful ways.

Drawing on various sources, the first version of these guidelines was developed in April 2023 by a working group that laid down general principles and specific instructions on using AI applications in teaching and learning. Since the world and our understanding of it are evolving rapidly, these guidelines serve as a preliminary agreement that may be influenced, for example, by national restrictions arising from data protection. In 2024, the university’s working group on AI in teaching updated the guidelines for using AI applications in thesis writing. As of September 2024, recommendations for acknowledging, describing and referencing the use of AI have been reviewed and unified.

1. The university encourages using AI (including chatbots) in teaching and learning to support education and develop students’ learning and working skills. When using AI applications, focusing on purposefulness, ethics, transparency, and a critical approach is essential.

2. In the context of a particular course, the lecturer has the right to decide how to use AI applications or, if necessary, limit their use. These instructions must be included in the course information in the Study Information System or clearly stated in the instructions for the specific assignment.

3. In the case of a written work, the use of a generative AI application must be appropriately described and referenced. Submitting a text created by AI under one’s name is academic fraud.

4. Personal data must not be entered into an AI tool without the person’s consent.

An AI chatbot could be compared to a companion who can be asked for advice anytime. However, it should be remembered that it is not a human and its output cannot replace evidence-based sources.

As with calculators, spell checkers, language-editing tools, search engines, and other similar tools, there is generally no point in prohibiting AI chatbots. Instead, one should consider how to learn to use them purposefully, ethically and critically.

By deliberately planning assignments that must be carried out with the help of an AI chatbot, it is possible to enhance students’ critical thinking, query-making, information evaluation, and problem-solving skills.

If a lecturer permits and promotes using AI chatbots or other AI-based applications in their course, they must consider whether all students have equal access to these applications. Students cannot be obliged to use a tool that requires creating an account with their personal email address. The differences between free and paid versions of AI chatbots must also be considered.

The study regulations specify that presenting a work produced by generative AI under one’s name without proper academic referencing or intentional use as part of one’s work without proper academic referencing is considered academic fraud.

If the work generated by generative AI is appropriately referenced, it is not considered academic fraud. If the lecturer prohibits using generative AI for an assignment, but the student still submits a properly referenced AI-generated task, it is not considered academic fraud, but rather a failure to meet the conditions for grading or passing the course.

The lecturer can make exceptions to the instructions for referencing generative AI in their course by indicating this information in the course information or the assignment instructions. No exceptions may be made to the requirements for referencing the source material used in the AI output, for which the regular rules for referring to written work apply. Therefore, the existence and content of the source material must be verified, citations are allowed to a motivated extent, and references to the source material must be correct.

In their article "Students and Copyright" (in Estonian), published in Juridica 2020/5, Aleksei Kelli, Age Värv, Merike Ristikivi, and Gea Lepik have written: "The Copyright Act allows works to be quoted to the extent justified. There is no single answer to what constitutes an 'extent justified’; rather, it must be assessed based on the specific circumstances of each case, according to what is reasonable for achieving the purposes of the quotation. Quoting a large portion or the entirety of a work constitutes generally reproduction (copying) of the work, which does not fall under the quotation exception."

AI chatbots can be used to support one’s studies; for example,

  • to work independently on tasks and assignments, to get explanations of concepts, to find ideas, to improve text, or to ask self-check and feedback questions;
  • to overcome writer’s block or the so-called fear of the blank page;
  • to serve as a brainstorming assistant;
  • to act as a programming aid;
  • to assist with editing and translating;
  • to develop critical thinking skills by evaluating the chatbot output;
  • to get a general overview of a large amount of material.

Lecturers can use AI chatbots to plan and prepare for lessons, facilitate work, and develop students' skills. For example, an AI chatbot may help lecturers save time when

  • creating and modifying teaching materials and presentations (adapting complex texts, providing examples appropriate to the specialisation);
  • drafting questions for a test paper, exam or self-check.

It is possible to develop students’ skills, for example, with assignments which students need to complete with the help of an AI chatbot. What matters is not the final result but the process, incl. writing effective prompts, evaluating the output, and holding a dialogue. Learners may also be asked to find an answer to a question of their choice with the help of an AI chatbot and have them write an analysis of the response.

If the lecturer wishes to restrict the use of chatbots, it is possible to

  • conduct an oral exam or a written exam with pen and paper in the classroom;
  • administer a written exam in a computer lab using the Safe Exam Browser in Moodle, configured to prevent access to other applications or browser windows during the exam;
  • reduce the weight of essay-type tasks in the final grade for written assignments or adjust the essay requirements, such as asking students to reflect on their personal experiences, opinions, connections to the specific material or data, or the Estonian context;
  • assign tasks that involve collecting original data through interviews, observation, fieldwork, archive studies, or other methods, and analysing the data;
  • use online tests for learners’ self-assessment rather than for grading and reduce their weight in the final assessment.

It must be clearly stated if the use of AI chatbots in a course or for assessment is prohibited.  Corresponding information must be clearly indicated in the course information or the assignment instructions. It must be kept in mind that using AI detection software to check students’ assignments is unjustified and not recommended, because (for now,) the working principles of AI detection applications are not clear and transparent.

For example, students can be informed of the restrictions as follows.

  • Using Microsoft Copilot, ChatGPT or any other AI-based software is not allowed in the course/quiz/test/exam.
  • Before you start completing the course assignments with the help of a fellow student or an AI chatbot (e.g. Microsoft Copilot, ChatGPT), please ask for my permission.
  • If you use an AI chatbot in a course where it is not permitted or do not refer appropriately to its use, it is academic fraud, which will be dealt with in the same way as other academic fraud cases.

These guidelines are a general advisory document compiled by the working group on AI in teaching at the University of Tartu. The purpose of the document is to assist the compilers of guidelines for authors of bachelor’s and master’s theses in updating the guidelines regarding the use of AI in theses. Each structural unit of the University of Tartu can supplement these guidelines based on the specifics of their faculty and discipline, as practices and limitations related to AI usage may vary across disciplines.

Below are some general principles for using AI applications, including chatbots, to compile theses. When composing theses, it is recommended to use AI applications that the Information Technology Office of the University of Tartu has approved. These versions ensure that the data used in conversations is protected and does not leak outside the university. More information can be found here.

  1. When composing theses, it is essential to recognise that the author bears full responsibility for the accuracy and quality of all information, research materials, analytical results presented in the work, and the correctness of references. When using AI applications, three fundamental principles should be followed: critical thinking, transparency, and ethics.
    While the use of AI applications during various stages of thesis writing is not prohibited, it is crucial to keep in mind that presenting AI-generated text in the thesis (and, more broadly, in any academic text) as one’s own thoughts constitutes academic fraud and is not in line with research integrity. For cases of academic fraud detection, warnings, reprimands, or exmatriculation may be issued to the student based on the decision of the relevant committee.
  2. Applications based on AI that do not create new content but only process existing text (such as translation programs like Google Translate, text correctors like Grammarly, and reference management tools like Zotero and Mendeley) can be used as supportive tools, and there is no need to cite them.
  3. Generative AI applications can be used, for example, 
  • to serve as sources of inspiration and as tools to develop one’s thoughts and ideas; 
  • to adjust the organisation and distribution of structural components within the thesis;
  • to identify synonyms for key terms or search phrases;
  • to find relevant sources or expand literature reviews;
  • to assist with translation;
  • to request explanations to understand complex concepts or theories;
  • to ask for feedback;
  • to detect programming issues.

Additionally, AI applications can be helpful when editing and formatting student-generated text during the final stages of thesis preparation. In such cases, there is no need to cite the applications.

  1. Using generative AI to gain an overview of a topic or to summarise information from various (foreign language) sources is not prohibited. However, when using the text in a thesis, the student must verify the content and existence of the sources and provide proper references for those sources. Additionally, it is essential to ensure that generated translations and used terminology are accurate and correct in content.

  2. Content created by a generative AI application cannot be presented under one's name without proper citation – doing so constitutes plagiarism. 

    AI applications must not be used for

  • creating lengthy text passages of one’s own text for a thesis (such as generating entire sections);
  • presenting unverified or false information;
  • fabricating or falsifying research data;
  • creating substantive arguments. 

Doing so constitutes academic dishonesty and a violation of research integrity.

It is not allowed to feed AI applications (research) data containing personal information, e.g. for writing a review or analysis of the data, or unpublished content from other authors.

  1. When using content generated by generative AI applications in a thesis, the same rules apply as to any other source. The content created by an AI application can be quoted or paraphrased in small amounts, provided proper citation rules are followed. 
  2. When generative AI applications are used to create substantial portions of a thesis, the methodology chapter must explain how AI was utilised and the application(s) used must be indicated in the list of references. For instance, the author should describe how an AI chatbot was used as an aid in composing text (example 1) or what substantive information was obtained from the chatbot, including the prompts used, the output received, and any modifications made to the output (example 2). The description of AI application usage should convey clearly the extent and the way it was applied in the text.

    Example 1. The assistance of the Microsoft Copilot (2023) chatbot was used in creating the structure of the survey used in this text.

    Example 2. The following definition is based on Microsoft Copilot’s response from 22 April 2023 to the question, “What is a language model?” The result was as follows: ‘[—]’ (Microsoft, 2023)
     
  3. In-text referencing depends on the citation style used (such as APA, Chicago, MLA) at the institute, department, or journal. In APA style, referring to AI applications follows the reference style for software. The owner/company of the AI application and the year of the version used must be indicated clearly in the in-text reference (example 3). 

    Example 3. AI applications also indicate that "while AI tools can assist in drafting texts, the main content and analysis must always come from the author" (OpenAI, 2024).
     

  4. When using images or other visual elements created by AI applications, the image’s origin must be acknowledged under the respective figure (example 4 and example 5). The application(s) do not have to be included in the list of references. 

    Double picture of black cat doing office work

  5. In the list of references, it is necessary to indicate the creator of the language model, the year of the used language model version, the specific application and its version, the type or description of the language model, and the application’s web address. For example, in APA style, a reference can be formatted as follows (example 6):

    Example 6

    Microsoft. (2024). Microsoft Copilot (March 3 version) [large language model]. https://copilot.microsoft.com/.
    OpenAI. (2022). ChatGPT (December 20 version) [large language model]. https://chat.openai.com/.

    If it is impossible to find all the necessary reference details, then the reference should include all the available information (example 7).

    Example 7

    JotBot. (2024). [Writing and research assistant software]. https://app.myjotbot.com/

If a student is uncertain about anything, it is strongly recommended first to discuss the matter with their supervisor(s).

Did you find the necessary information? *
Thank you for the feedback!
About text formats