Writing with AI

The University of Graz has decided not to prohibit the use of AI technologies. However, the University's orientation guidelines stipulates that lecturers will determine "which text-generating AI systems may be used and in what way" (Vice-Rectorate for Studies and Teaching, 2023, p. 3). Thus, the use of these technologies may be permitted in some courses but not in others. It is up to the lecturer, based on didactic considerations, to determine how AI technologies may be used by students and how the usage must be documented and declared.  

It is therefore important that you discuss and clarify the use of AI technologies in your writing processes with your lecturers for each course and for each purpose.

Last updated: 08 April 2024

First and foremost, it is important to discuss the use of AI technologies with your lecturers. In general, AI technologies can be used for various tasks such as revising research questions, working with literature, developing outlines, or editing texts stylistically (see orientation guidelines). 

To ensure that AI technologies benefit your learning, we recommend that you ask yourself the following questions before using AI tools for writing(see also Questions for Reflection on Usage of AI in Writing):

• What is the specific purpose of using AI technologies in my writing or writing process?
• Is this purpose permitted in the relevant course context?
• If so, how should I document, label, and reflect on my use of AI technologies?
• Which AI tools are suitable for the task at hand?
• How could I perform this step without the use of AI technologies?

Here you can find instructions for several AI tools. Additionally, this resource collection provides more information on the functioning, potential, and limitations of generative AI technologies.

Last updated: 08 April 2024

When using AI technologies, it is important to follow the principles of good academic practice and discuss their use with your lecturers. A critical and reflective approach to these technologies is essential to assess and observe their limitations and potential. The information provided on this page can serve as a basis for orientation.

It is crucial to consider the following points when using AI in your writing:

• Use AI only according to the AI guidelines of the respective course
• Label and document AI use
• Check the content of all AI-generated information
• Know about and consciously deal with the limitations and weaknesses of AI systems
• Consider data protection at all times (of your personal data and, if applicable, research data)
• Comply with the principles of good academic practice
• Ethical considerations (see, for example, this article on principles for the ethical use of artificial intelligence)
• Responsible use of AI technologies preserving your autonomy as an author and your independent contribution to the text

Last updated: 30 October 2023

Under no circumstances should you use AI technologies...

• ...in such a way that their use contradicts the requirements of the respective course.
• ...without transparently documenting their use.
• ... to transfer responsibility for your work to an AI system: As the author of a writing assignment (e.g., an excerpt, a seminar paper, or a master’s thesis) submitted for assessment, you always carry responsibility for your text. You must not delegate this responsibility to an AI system.
• ...in such a way that it impairs your independence as an author (see question on independence) or contradicts the guidlines of good academic practice.
  
• ... in a way that it obstructs your learning process: While utilizing AI technologies, it may be tempting to view them as a means of relieving you from tedious writing tasks and saving time. However, this assumption is only partially true, as editing an AI-generated text can be time-consuming. It is also important to consider that relying solely on AI support may result in missing out on valuable learning opportunities and acquiring limited skills during your studies. To ensure that the use of AI technologies is beneficial for your learning, we recommend to ask yourself the following questions before using AI tools for writing:
  • What is the specific purpose of using AI technologies in my writing or writing process?
  • Is this purpose permitted in the relevant course context? 
  • If so, how should I document, label, and reflect on my use of AI technologies? 
  • Which AI tools are suitable for the task at hand? 
  • How could I perform this step without AI technologies?

Here you can find further questions for reflection on the use of AI in writing. 

Last updated: 31 January 2024

If you decide to use generative AI technologies, it is important to be aware of their weaknesses and limitations. The following is a brief overview of their most significant limitations and weaknesses:

• AI technologies are often referred to as 'stochastic parrots'. They are based on statistical models and generate text sequences that try to mimic the patterns and structures of the training data. These texts can appear sophisticated and may give the impression that the technologies are ‘intelligent’. However, it is important to remember that they are not capable of real intelligence. As AI systems generate probable word sequences without the ability to verify their content, argumentative validity, or factual accuracy, they are often called 'stochastic parrots' (Bender et al., 2021).
• AI technologies may produce false content and citations. AI-generated texts do not prioritise factual accuracy. For instance, the current version of ChatGPT does not have the capability to verify the accuracy of the information it produces. As a result, the generated texts may contain errors and incorrect information. Moreover, the stochastic nature of these technologies often leads to the creation of 'hallucinated' facts and sources that are 'probable' on a linguistic level, but not necessarily factually correct.
• AI technologies do not provide up-to-date information: AI systems are based on training data that originates from a specific time period. For example, the current version of ChatGPT only takes into account data created up to September 2021 and therefore cannot consider more recent information and events.
• AI-technologies are subject to bias: In addition to the temporal limitation of the training data, the selection of particular kinds of training data can lead to biases and discriminatory effects (Bender et al., 2021). Additionally, the generated texts often reflect hegemonic positions, resulting in an underrepresentation of marginalized views. For more information on bias and fairness in AI systems, please refer to this website.
• AI technologies may lack contextual knowledge. Despite being based on large amounts of data, AI models often have a limited understanding of context. This lack of semantic understanding can lead to the generation of texts that are factually accurate but not suitable for the intended communicative purpose of the user.
• AI technologies commit plagiarism and copyright infringements. It is often unclear to users which sources the generated texts are based on. This can result in the inclusion of longer text sequences in the generated text without labelling them as direct quotes, which is a form of (plagiarism. Thus, if you incorporate generated texts directly into your work without verifying sources and potential copyright infringements, you may inadvertently violate the guidelines of good academic practice.

These limitations require careful examination of the content of all generated information and texts (see also the question on the use of AI-generated texts in your work). For more information on the functionality, potential and limitations of generative AI technologies, please refer to these resources.

Last updated: 30 October 2023

No, a generative AI like ChatGPT cannot write your texts for you. If you submit a writing assignment for assessment, you as the author are always responsible for the text. This means that you must be able to explain what you have written, why you have written it, and how. Delegating this responsibility to an AI system is not acceptable.

Even if it is permitted to use AI technologies for particular purposes in a course, AI must be used responsibly, which means:

• Always follow your lecturer's guidelines regarding the use of AI technologies in a particular course. Using AI technologies in writing assignments, theses, or other examinations in violation of your lecturer's guidelines may be considered academic cheating and result in a negative assessment.
• Ensure that there is an individual contribution in all your writing assignments. The disproportionate use of AI tools can undermine your ability to make an individual contribution to your assignments, as discussed in the question on independence.
• All writing assignments must comply with the guidelines of good academic practice. This includes, among other things, the documentation of the sources and tools used in the writing and research process. Thus, you must indicate which technologies you have used as tools in your work process. If you use generative AI technologies, a further problem may arise in relation to the guidelines of good academic practice: The tools currently available do not make it possible to trace reliably on which texts the generated texts are based on. Therefore, AI-generated texts may contain unmarked direct quotations (plagiarism) from other sources. When submitting a thesis for assessment, the University of Graz requires students to submit an academic integrity statement. This document states that the thesis contains original work of the student and affirms that it was written in accordance with the principles of good academic practice, without any undeclared external assistance. Additionally, all sources used must be correctly labelled (see instructions for submitting a BA thesis).

Attention! If you utilize AI technologies in your work, ensure that they align with the guidelines outlined in this document. Failure to do so may be considered academic deception and result in a negative assessment (see § 25 para. 6, § 37 para. 4, § 38 para. 8, § 39 para. 9, § 40 para. 1-4 Statutes of the University of Graz, Study Regulations).

Last updated: 30 October 2023

It is important that you use AI technologies in a way that ensures you make an appropriate personal contribution to your texts. In academic writing, this is demonstrated by how you incorporate existing ideas, materials and literature into your texts to develop your own argument. In addition, students are expected to work independently, which means that they should be able to complete assignments on their own and only use designated and documented tools and resources. When submitting a thesis for assessment, the University of Graz requires students to submit a declaration (in German ‘Eigenständigkeitserklärung’). This declaration states that the thesis contains the student’s original work and affirms that it was written in accordance with the principles of good academic practice, without any undeclared external assistance.

Lecturers assess whether or not students demonstrate the ability to work independently and make an adequate personal contribution to their texts. In this context, it is crucial to note that the importance of making a personal or original contribution in a text increases in the progression from undergraduate to graduate studies and doctoral research and can thus vary greatly depending on the type of writing assignment.

However, it is difficult to provide concrete information about which use of AI technologies violates these requirements and which does not. The University of Graz has not (yet) formulated any general criteria that could be used to assess this requirement in the context of AI-supported work processes. Rather, it is up to the faculty to determine and verify whether the use of AI for a particular purpose is compatible with these requirements.

Given this situation, it is especially important that you always discuss the use of AI technologies with your lecturers. In addition, you are obliged to be transparent about how you have specifically used the technologies (e.g., by documenting the prompts, reflections or drafts used).

A German legal opinion on the use of AI software in the university context contains some examples of the assessment of students' personal contribution to their texts that illustrate the problem well, but cannot provide a universally valid answer to this question:

“If AI-generated texts containing content relevant to an assignment are copied without significant intellectual input from the student, these assignments can probably not be considered ‘independent writing without outside help’, as there is no longer independence. In this context, the concept of independence means that the results must be based on the student's own knowledge, i.e., the knowledge required to complete the assignment must come from the student. If this knowledge is ‘generated’ by the AI, the work can no longer be considered independent.
This means that, depending on how it [AI] is used, the resulting text may be considered a case of cheating or deception. If, for example, students use ChatGPT or, in perspective, a ‘Word AI’ and copy the knowledge it generates and formulates word for word–without making this transparent and without having significantly steered the text production– an attempt to cheat must be assumed. By contrast, the use of an AI may be permissible if the AI merely suggests additional wording without relieving the student of the task of completing the assignment. The inspirational use of an AI as a thought-provoking stimulus may also be considered permissible.
The permissibility of the use of an AI is questionable if, for example, the student uses knowledge relevant to the completion of the assignment to formulate their prompt, and if the AI product is then included in the examination performance. In all cases, the relevant examination regulations, the declaration of independence, and the examiner's assessment of the extent of the student's intellectual contribution will always be decisive. Thus, evaluating students’ independence depends primarily on the individual case and possibly also on the requirements of the individual departments". (Hoeren 2023, p. 33, translation by Writing Centre)

Last updated: 30 October 2023

We strongly discourage copying AI-generated texts into your own work. This is because you as an author are always responsible for the text when you submit a writing assignment (e.g., an excerpt, a seminar paper or a master's thesis) for assessment.

It is therefore essential that you critically evaluate AI-generated texts before using them in any way. Especially because of the error-prone nature of AI technologies, you should carefully verify all information (see question about the weaknesses and limitations of AI) and be sure to include other scholarly sources in your research.

AI-generated texts may appear stylistically polished and convincing at first glance, but upon closer examination, they often fail to fulfill their intended communicative purpose or distort the intended meaning. Therefore, it is essential to revise AI-generated texts to ensure a good fit in terms of content, structure and style. For more information on text revision, you may also refer to this this video on revising texts. The University of Mannheim recommends the following procedure specifically for for dealing with AI-generated texts:

• Checking the content for:
  • Facts, misinformation, errors: is the information provided factually correct?
  • Specific or generic answers: Is the question answered specifically enough or is the answer generic and not connected closely enough to your specific purpose?
• Checking the perspective: What is the cultural and social perspective embodies by the generated text?
• Checking the fit: Does the text match your expectations and your intended purpose?

The orientation guidelines of the University of Graz describes the handling of direct quotations as follows:

„The adoption of AI-generated text passages should be treated similarly to conventional citations, with the specification of the AI system and the interaction with the technology. This ensures objectivity and transparency in the use of AI-generated text”. (Vice-Rectorate for Studies and Teaching, 2023, p. 2)

In an APA citation style this can look like this (see this APA article):
In-text Citation:
Your own text "direct citation by ChatGPT" (OpenAI, 2023).
In the bibliography:
OpenAI. (2023). ChatGPT (version of March 14) [Large language model]. https://chat.openai.com/chat

Another potential issue with AI technologies is the possibility of plagiarism and copyright infringement. It may not always be clear to users which sources the generated texts are based on, leading to the inclusion of unmarked plagiarism. To avoid violating the principles of good academic practice, it is important to properly cite any generated texts that are incorporated into your own writing.

Finally, it is essential to adhere to the specifications of the respective course when using AI technologies. If you directly adopt text passages, it is essential to transparently document your use of AI and maintain your independence as an author. Please refer to the question independent working when using AI for further information. 

Last updated: 30 October 2023