© Adobe Stock Datei NR.: 275810514 Bildungseinrichtungslizenz von Feodora

The progressive development of AI technologies opens up a variety of opportunities in the field of teaching and learning, but also brings ethical, social and legal challenges. This page highlights the good practice in dealing with generative AI at TU Wien in the area of study and teaching, taking into account a university didactic approach. The decision whether teachers use AI systems is up to them. If they do use them, the guidelines must be clearly defined and the focus must be on how to deal with them.

The responsibility for the results of such generative systems always lies with the human being and they must therefore be critically questioned.

© Adobe Stock Datei NR.: 618444229 Bildungseinrichtungslizenz von Feodora

Interdisciplinary integration: If possible, AI topics should be integrated into courses so that students learn how to deal with them. AI permeates many disciplines and opens up different opportunities to act as a learning partner, for example. 

Practical relevance in education: The integration of generative AI should focus on project-based learning in which students actively use AI. This promotes practical knowledge and creative problem solving.

Ethics and society: When using AI in teaching, attention must also be paid to ethical and social aspects of AI. Raising students' awareness of responsibility and the impact of AI on society is a key issue.

Bias: Generative AI systems are never free of bias because they reflect the bias of the materials they are trained with; this bias can harm individuals, groups of people, or society as a whole. Appropriate caution is therefore required when dealing with synthetic media; students must therefore be encouraged to think critically in this context as well.

Collaborative learning: Collaborative learning and sharing of ideas and experiences between students can be encouraged. AI's such as ChatGPT et al. can be useful as a learning partner to promote critical thinking, as the results are often flawed.

Privacy: in the context of AI systems (as with other digital systems), privacy can be an issue when personal data is used to train the AI or when the AI is used to analyze personal data. AI systems must also comply with data protection standards such as purpose limitation of personal data or data transfer to a "third country" in accordance with data protection requirements.   Students must be made aware that everything that is entered is also made publicly accessible accordingly.

Transparency and explainability: The AI models and algorithms used should be as transparent and explainable as possible. To this end, reference can also be made to current research results on different systems. This is important to create trust in the decisions of the AI systems and to prevent discrimination.

Responsible research: AI research should be conducted ethically and responsibly. Potential risks and impacts on society and the environment must be considered.

Respect intellectual property rights: intellectual property rights must be respected in the development and use of AI systems. To this end, it may be helpful to clarify the rights to data, models, and software in advance.

Support low-barrier access to teaching and learning materials: Common generative AI applications can also be used to generate image or video descriptions, or also to enable automated captioning of audio and video materials. This can promote low-barrier access and support equity. See also:

Barrierefreiheit in der Lehre

Responsible research: Conduct AI research ethically and responsibly. Consider potential risks and impacts on society and the environment.

Respect intellectual property rights: Ensure that intellectual property rights are respected in the development of AI systems. Clarify rights to data, models, and software up front.

Handling AI in Written Work: Since Large Language Models such as ChatGPT can produce entire written papers (e.g., bachelor's theses, seminar papers, etc.), as well as presentations, it is important to focus on the process of producing such papers. ChatGPT et al. must be documented as a tool and also the places where AI's were used must be marked accordingly. Since responses are often incorrect, it is important that students learn how to use them.

In particular, it is important to document what generative AI systems were used for in the creation of the paper, how it was ensured that the media thus generated was correct in content. As an appendix, it is also useful to document what was learned from it. For substantial contributions, indicate the generative AI system used, as well as the prompts and the date of the query used to generate the media.

Generative systems must be documented as aids and also the places where AI's were used must be marked accordingly.

For more informations about rules for plagiarism look into: Leitfaden zum Umgang mit Plagiaten in studentischen Arbeiten an der TU Wien, opens a file in a new window .

If generative AI texts are used in one's own work without substantial changes, they should be treated similarly to citations from other sources. This means that it must be clearly marked which text passages were created by a generative AI system.

In addition to identifying the synthetic material, a source citation is required that includes the "prompt" used to generate the media, a date, and the system used. ("Prompt example, 09/12/2023, ChatGPT, version").

Collect and share experience values: In order to be able to continue to progress our learning path in a goal-oriented manner as TU Vienna, the experiences you have gathered are very important to us and will be gladly read. Experiences can be sent to anna.fuessl@tuwien.ac.at.

Academ. Integrity: students must learn that any use of generative AI must be clearly and transparently labeled, the prompts used must be disclosed, and links to the results must be provided. It is important to understand that failure to disclose the use of generative AI in a student paper is tantamount to plagiarism.

"Cognitive Bias": The linguistic fluency of systems like ChatGPT comes across as Intelligent and Persuasive, thus influencing perception. This is true even if the written information is incorrect.

AI and the environment: On the level of sustainability, training of AI systems is very energy-intensive and correspondingly burdensome. This applies both to established systems such as ChatGPT and to models that are to be trained themselves.

Discriminatory algorithms: Avoid using algorithms that could provide discriminatory results based on gender, skin color, or other protected characteristics.

Uncontrolled autonomy: Avoid using AI systems with high autonomy, especially when human supervision is required. The responsibility should always lie with the human.

Lack of data protection measures: Ensure that adequate measures are in place to protect personal data. Inadequate safeguards can lead to data breaches.