University of Gottingen AI Policy

Against the background of experiences gained since 2023 with AI models in study and teaching, the university reinforces its recommendation to integrate these tools into teaching as extensively as possible and to promote their use by instructors and students.

Even though the use of AI is generally encouraged, it can sometimes be pedagogically useful to adopt more differentiated approaches. Students must, in any case, have clarity about the rules for AI use in courses. This should be discussed with participants at the beginning of each semester in all courses.

For examinations taken without supervision (e.g., term papers, essays), students have access to AI text models. Prohibiting their use is not advisable, not only because supervision is lacking, but also because meaningful use of such tools is already part of prior school education and will be important for future professional life. Proper use in the sense of good scientific practice (GSP) should therefore also be part of the examination.

It should also be noted that different disciplines or programs may have specific requirements for AI use and its declaration. While in some disciplines the declaration may, for example, be provided solely as answers to these questions in an additional document attached to the work, in other disciplines it is customary to include specific information in the methodology section of the work. Instructors will provide details on the exact requirements in the courses.

At the same time, independent academic performance must be ensured.

• In supervised exams (e.g., written exams, oral exams), AI tools are not permitted unless explicitly allowed.

• In unsupervised exams:

1. AI models should generally be permitted as aids.

2. Their permissibility ends where AI-generated content is not transparently declared.

3. AI use must be transparent. Students should explain how AI tools were used (see Appendix 2).

4. Alternatively or additionally, a methodology section may be introduced in which all tools used are described.

5. If AI competencies are not a direct assessment objective, transparent AI use should neither positively nor negatively affect grading.

6. Examiners should adjust examination formats and questions so that the independence of student work can still be verified even when AI tools are used. For suitable exam formats, examiners may decide to waive the declaration requirement.

Against the background of experiences gained since 2023 with AI models in study and teaching, the university reinforces its recommendation to integrate these tools into teaching as extensively as possible and to promote their use by instructors and students.

Developing competence for the critical and responsible use of AI systems must become an integral part of teaching at our university.

Existing AI systems also significantly increase students’ self-responsibility for their own learning success. This can, for example, be supported by instructors providing opportunities for regular voluntary learning assessments (without AI).

The GWDG offers a portfolio of AI services for all university members (students and instructors), e.g., its own chatbot (ChatAI). This chatbot provides access to AI models hosted by the GWDG as well as to external models such as those from OpenAI (ChatGPT). These offerings should be introduced and explained to students of all disciplines at the beginning of the semester.

Great course to get started with AI, without prior knowledge. Also includes prompts, ideas for responsible use and tips for your learning process.

In the higher education context, it is especially significant that these models can generate texts of high quality in a question-and-answer format. This also applies to the generation of source code for programs.

The recommendations presented below therefore explicitly apply both to traditional texts and to program code.

Sie benötigen KI-Dienste, die Datenschutz, Sicherheit und wissenschaftliche Anforderungen ernst nehmen. Gleichzeitig wollen Sie moderne KI nicht nur im Chat nutzen, sondern auch über APIs, für dokumentenbasierte Assistenten, Code-Unterstützung, Bildgenerierung, Sprachverarbeitung und weitere spezialisierte Anwendungen.

With an Academic ID, you can also request an API key via the KISSKI website (click on “Book” on the page). You can use the API key in another webfrontend, plugin or your own code.

It is therefore advisable to limit the use of text-generating AI tools to specific tasks and purposes, such as generating text introductions or designing transitions.

The “mental and manual work” involved in scientifically sound research, reading and excerpting texts, and citing and writing scientific texts in accordance with scientific quality criteria has to be the initial focus.

We, the students, take responsibility for the texts we produce and clearly indicate where and for what purpose we have used AI tools in the writing process.

Proper use in the sense of good scientific practice (GSP) should therefore also be part of the examination.

In the academic context, it is additionally relevant that LLMs not only fabricate facts but may also “invent” sources and present them in the same format as real references. Listed literature references — including author initials, journal titles, volumes, page numbers, and similar details — may therefore be entirely fictitious. Users therefore bear particular responsibility for how texts generated by LLMs are used.

We, the lecturers, refer to the university-wide principles of good academic practice.

We, the students, take responsibility for the texts we produce and clearly indicate where and for what purpose we have used AI tools in the writing process.

A prerequisite for this is transparency: all participants must disclose the extent to which LLMs were used.

Consistent culture of transparency: The use of ChatGPT should be openly communicated by both instructors and students.

AI use must be transparent. Students should explain how AI tools were used (see Appendix 2).

Alternatively or additionally, a methodology section may be introduced in which all tools used are described.

We, the students, take responsibility for the texts we produce and clearly indicate where and for what purpose we have used AI tools in the writing process.

Appendix: Declaration on the use of ChatGPT and similar tools in exams

In this paper, I have used ChatGPT or other AI tools as follows:

Texts generated by AI chatbots are so similar to those written by humans that their use can hardly be detected through automated identification methods such as plagiarism software.

This template facilitates the documentation of academic plagiarism in student texts. You may use this template to share instances of plagiarism with the Examination Office, which will take over from there.

Hier finden Sie die in der Handreichung genannte Vorlage zur Plagiatsdokumentation , welche zur Dokumentation von Plagiaten in schriftlichen Prüfungsleistungen sowie zur Meldung dieser Plagiate an das Prüfungsamt der Sozialwissenschaftlichen Fakultät verwendet werden kann.

Against the background of experiences gained since 2023 with AI models in study and teaching, the university reinforces its recommendation to integrate these tools into teaching as extensively as possible and to promote their use by instructors and students.

Students must, in any case, have clarity about the rules for AI use in courses. This should be discussed with participants at the beginning of each semester in all courses.

We, the lecturers, refer to the university-wide principles of good academic practice. Furthermore, we set out the assessment criteria in general and expectations regarding AI tools in particular at an early stage in the courses for which we are responsible.

Examiners should adjust examination formats and questions so that the independence of student work can still be verified even when AI tools are used.

The GWDG offers a portfolio of AI services for all university members (students and instructors), e.g., its own chatbot (ChatAI).

the competence to assess relevant aspects of data protection, personal rights, and copyright when using AI, as well as an understanding of the difference between using self-hosted models at the GWDG (e.g., chat-ai.academiccloud.de) and external models (e.g., ChatGPT from OpenAI),

The GWDG offers a portfolio of AI services for all university members (students and instructors), e.g., its own chatbot (ChatAI). This chatbot provides access to AI models hosted by the GWDG as well as to external models such as those from OpenAI (ChatGPT).

Here you will find the university's own ChatAI, ImageAI, and VoiceAI.

In order to use the models hosted by the GWDG, the user’s input/requests are processed on the GWDG’s systems. Protecting the privacy of user requests is of fundamental importance to us. For this reason, our service in combination with the self-hosted models does not store the contents of the requests (chat history), nor are requests or responses stored on a permanent memory at any time.

In order to use the OpenAI models, we send the respective request (user input) from our server to the Microsoft servers (external service provider).

Information about the users themselves is not forwarded by GWDG. However, the user’s enquiry is forwarded unfiltered, i.e. personal information contained in the enquiry itself is forwarded to the external service provider.

Not all of these may be approved for use by your institution. In order to restrict the use to certain models, the authorized person of your institution must inform us in writing of the selection of models.

Against the background of experiences gained since 2023 with AI models in study and teaching, the university reinforces its recommendation to integrate these tools into teaching as extensively as possible and to promote their use by instructors and students.

Developments observed in the context of Wikipedia have also shown that the use of such systems cannot and should not be prevented or prohibited. On the contrary, the University of Göttingen understands AI models as tools that will be used and for whose intelligent application students and instructors must acquire competencies.

The following recommendations can only be provisional due to the rapid pace of development. Language-based AI models will sooner or later have significant influence in many areas of society, which makes it essential for us as a university to repeatedly engage with both the opportunities and risks involved.

Developing competence for the critical and responsible use of AI systems must become an integral part of teaching at our university.