Navigation auf uzh.ch

Suche

Recommendations on the Use of Generative Artificial Intelligence at UZH

AI-generated image of UZH, generated in Deep Dream Generator based on a photograph of the main building.

Use of artificial intelligence in Teaching

UZH has a broad range of teaching and learning formats. The way in which generative artificial intelligence AI can be used in the various formats varies widely depending on the discipline and subject matter.

Responsibilities

The faculties and the individual study program directors are responsible for drawing up specific guidelines and policies. The topic of generative AI should be considered when designing the overall curriculum and the necessary measures taken. Those responsible must ensure that students are aware of whether and to what extent generative AI may be used in assessments. If the use of generative AI is not or only partially permitted, they should ensure the relevant assessments are conducted fairly.

Key questions

Based on the guiding principles of the Extended Executive Board of the University, and the guidelines or policies that apply to the faculty or relevant study program, the following key questions will help teaching staff understand how they can integrate generative AI tools in their teaching and learning formats:

  • To what extent do the existing learning objectives for the modules or individual lectures need to be adapted?
  • How can students be taught to engage with generativeAI tools in a sensible, competent and critical way?
  • What didactic opportunities exist to actively integrate generative AI tools?
  • Which aspects of media literacy can students be taught through the active use of  generative AI tools?
  • How are generative AI tools used in potential occupational fields? How can students be best prepared for this?
  • How do existing assessment formats need to be adapted to reflect the potential use of generative AI?

Specific ways of actively using generative AI tools such as ChatGPT in teaching and assessments are available on Teaching Tools. In any case it is important to ensure that all involved know whether and how the use of generative AI tools is permitted in each particular context and the consequences of violating the relevant regulations.

Teaching staff may not require students to use generative AI tools that request personal data to log in or register or that entail costs during their courses. This is in line with the principle of equality of opportunity in teaching and learning and in assessments.

It is also the responsibility of students to actively inform themselves of how and to what extent generative AI tools may be used as aids in each individual course.

Didactic resources

Helpful didactic resources on generative AI can be found on the UZH Teaching Tools website:
Teaching Tools: Didactic Resources on generative AI at UZH

Declaration of authenticity when using generative AI tools

A declaration of authenticity can be used for written assessments to confirm, among other things, that the tools and sources used have been correctly cited and referenced. Depending on the subject, generative AI tools are deemed aids and would therefore fall under the scope of the standard declaration of authenticity.

If, however, subject-specific features mean it would make sense to include explicit information about the use of generative AI tools, this can be added to the existing declaration of authenticity.

UZH Legal Services and Data Protection has summarised the legal aspects that must be borne in mind below.

  • The person completing the declaration must confirm that they have only used the permitted aids or tools. Due to the novelty of the technology, it is worth clarifying that generative AI tools are aids.
  • It should be made clear that the person making the declaration bears full responsibility for the use of a generative AI tool. In particular, the use of generative AI tools can lead to copyright infringements and violations of data protection law.  
  • Any use of generative AI tools must always be indicated.
  • The common wording that was previously used to confirm that someone’s work was written “in their own words” should be adapted if generative AI tools are permitted as aids. 
  • If necessary, a note can be included stating that violations of the declaration of authenticity have legal consequences. On account of the novelty of the technology, however, legal assessments with regard to generative AI tools should be avoided.

Taking into account these points and the relevant learning objectives, the following wording could be added to a declaration of authenticity to address generative AI tools: 

“I hereby declare that I carried out this work independently and without the use of unauthorized aids (generative AI tools are also aids). I am aware that I bear full responsibility for the selection (delete if specifications were given on generative AI tools), adoption and results of the AI-generated outputs I used. I also declare that all use of generative AI is fully disclosed. I have listed the generative AI tools used along with their product names in the list of aids and resources (modify if there is no list of aids and resources).

I acknowledge that work that violates the principles set out in this declaration of authenticity may have legal and disciplinary consequences.”

Citing and referencing AI-generated content

A central and generally-accepted principle of good scientific practice is transparency and reproducibility when describing research methods and results in scientific or academic works and publications. This includes the use of generative tools.

Typically, generative tools have to be declared, while non-generative AI-based tools, such as a spell check in Microsoft Word, do not.

Depending on the subject area, the relevance of a specific tool or aid may vary and a declaration may be advisable (e.g. the difference between using a translation tool in a language course versus in a mathematical context). If a declaration requirement is sought, it helps to be guided by different levels of intended use.

When referencing generative tools, it is advisable to be guided by common referencing formats, such as the American Psychological Association (APA) or the Modern Language Association of America (MLA). The two examples below show the use of both formats to directly quote texts generated by ChatGPT.

The question of which citation method is suited to which purpose and subject area is ultimately down to the discretion of the study program coordinator or instructor. In any case, it should be ensured that students are actively informed of the requirements.

Citing ChatGPT from OpenAI according to APA style

APA format

Author (year). Title (date and version) [description of the model]. URL
APA bibliography entry, example OpenAI (2023). ChatGPT (May 2023 version) [large language model]. https://chatopenai.com

APA in-text Zitation, Beispiel

«Zitat» (OpenAI, 2023)

American Psychological Association: How to cite ChatGPT

Citing ChatGPT from OpenAI according to MLA style

MLA format

“Prompt text” prompt. Title, day month version, author, day month year, chat.openai.com
MLA bibliography entry, example How do I cite ChatGPT in MLA-Style?” prompt. ChatGPT, May 2023 version, OpenAI, 23 July 2023, chat.openai.com
MLA in-text citation, example (“How do I cite ChatGPT in MLA style?”)

MLA Style Center: How do I cite generative AI in MLA style?

Citing possibilities for the use of ChatGPT and other generative AI tools in continuous text, based on the APA format:

  • Direct citation of passages: “Quote” (OpenAI, 2023)
  • Paraphrasing of passages: Paraphrased text (OpenAI, 2023, paraph.)
  • Translation of a passage using ChatGPT: Passage (OpenAI, 2023, trans.)
  • Translation of a passage using DeepL: Passage (DeepL SE, 2023, trans.)
  • Stable Diffusion image generator: prompt in caption: Fig. 1: “University of Zurich in space”; Stability AI, 2023)

If generative tools are used for more than one purpose, such as direct citation, paraphrasing or even translation, it may make sense to reference these purposes accordingly in the running text and in the bibliography in the form of an index of aids or resources in the form of a table or list.

Citation in scientific publishing

For scientific publications, the specific guidelines and policies of the relevant journal or publisher with regard to AI should always be borne in mind.

For example, Nature and Elsevier have defined regulations on dealing with authorship and AI-generated content (e.g. text, figures, images):
Nature: Editorial Policies
Elsevier: AI Policies