AI in studies and teaching

/

• What is artificial intelligence (AI)?

  AI is intelligence realised by machines or programs. We usually differentiate between weak AI (or narrow AI) and strong AI (or artificial general intelligence). While weak AI can solve clearly delimited problems, strong AI (which could not really be implemented previously) shows general intelligent abilities and can thus solve any task requiring intellectual capabilities. The more popular, successful AI tools are rather weak AI, such as computing programs defeating human world chess champions or world go champions or programs generating images from text prompts. Also language models, such as GPT are rather referred to as weak AI (for an exemplary description of capabilities of GPT 4 see, for example, Bubeck et al., 2023). For a broader discussion on the capabilities, risks and opportunities of modern language models, please see, for example, Bender et al., 2021; Brown et al., 2020.

  AI systems are available in many different forms and complete a variety of tasks. There are AI systems that can play various board games and video games. Moreover, there are AI systems that support physicians in diagnosing diseases (expert systems), that identify objects in images, sort waste automatically, decide on credit allocation, drive cars independently, etc. Recently, especially one special category of AI systems, the so-called language models, such as Chat-GPT have attracted attention and have increasing significance.

   

   
• What AI tools are available?

  There is no answer to this question since new and improved tools enter the market each day. While language models including a chat function dominated the first phase of the hype, image and video generators are now available as well and are developing extraordinarily. 

  A brief overview of which AI tools are on the market in the area of study and teaching can be found on the website www.vkkiwa.de/ki-ressourcen/
• May I use AI tools for my studies?

  Frankly speaking, there is no easy answer to this question. For example, students of Computer Science are developing and experimenting with language models themselves. Students of Translation Studies are using AI for translations. Every text generation program contains little helpers in the form of AI. Students enrolled in teacher education programmes learn how school students prepare homework assignments and presentations and how to deal with the use of AI in this context. And many courses examine the impact of AI tools on society, culture and technology from an academic perspective.

  Of course, exams and academic theses must clearly demonstrate what YOU know and are able to do. We do not hold exams to assess AI tools.

  Therefore, teachers and examiners in each discipline decide whether and how students may use AI tools as aids. For this purpose, the University of Vienna has provided guidelines for our teachers. A crucial aspect in the context of AI is transparency. Teachers have to specify permitted materials before the course and/or exam. In view of the wide range of available tools, this is not an easy task.
• What can I do if teachers do not provide any information?

  For the sake of clarity, ask teachers in the first course unit or before the exam whether and in what form you may use AI tools.
• What do I need to consider when using AI tools?

  Whether you use AI tools in private life or for your studies, you should consider a few questions and topics:

  • What data does the AI tool want me to provide and am I prepared to share these data (especially personal data)? What are the interests of the company providing the AI tool? Never enter personal data or secrets in an AI tool! This applies to both your personal data and the personal data of others.
  • On which basis was the AI tool trained? Can I be sure that the output is correct, complete and especially unbiased? When using AI tools, you are always responsible for how you use the results. Therefore, checking the output of AI tools is an important task. AI tools can hallucinate. This means that they may present completely fabricated texts as the truth or as reliable, existing sources. Be careful!
  • Document the prompts you entered and when you entered them, the results that the AI tool generated and the way in which you used these results.

  Example: using ChatGPT in a safe way 

  Large language models, such as Chat-GPT, do not have a source of truth due to their network architecture, as Open AI describes in its blog article at the beginning of the section on 'Limitations'. “ChatGPT sometimes writes plausible-sounding but incorrect or nonsensical answers. Fixing this issue is challenging, as: (1) during RL training, there’s currently no source of truth; (2) training the model to be more cautious causes it to decline questions that it can answer correctly […]“. Therefore, we have to emphasise again that chat bots are no knowledge machines in any case. Quite the opposite: In the end, it is always the obligation of the human using the AI tool to check the accuracy of the output and to assume responsibility for the result. A flow chart by Aleksandr Tiulkanov illustrates this in an impressive way:

  Illustration: Aleksandr Tiulkanov (© https://creativecommons.org/licenses/by/4.0/deed.de)

  
• What changes with regard to writing academic theses?

  When writing an academic thesis, please discuss with your supervisor in advance which aids you may use and how you have to document in a transparent way that you completed the achievement independently. This may also differ between disciplines. In any case, you should discuss the following aspects (your supervisors are aware of this):

  Discuss with your supervisor how you correctly approach academic research and writing.

  For your academic thesis to be assessed as an independent achievement, you must

   

  • have documented any aids used

   

   

  • in the section where the aids been used and

   

  • in the description of the thesis’ methods;

   

   

  • have cited intellectual property of other persons according to the disciplinary rules and referenced these in the bibliography;

   

  • have the right to use copyrighted images and media in the thesis;

   

  • have documented all (raw) data, protocols and analyses generated during preparation in a transparent way and accessible at any time;

   

  • have been transparent about any texts and images generated by (AI) tools and their adjustments throughout the thesis preparation process;

   

  • have explicitly mentioned any contribution to the content by third parties (such as data processing, analysis, etc.) and have acknowledged these persons accordingly (e.g. in the acknowledgements);

   

  • did not make use of any impermissible support regarding content (e.g. ghost writing) in addition to the approved supervision;

   

  • have mentioned any content overlaps with their contributions in courses (e.g. bachelor’s paper, seminar paper).

   

  Discuss with your supervisor how you best approach your thesis project. You will also have to confirm the aspects above when submitting your academic thesis. In case of the suspicion that your thesis was not an independent achievement, a study law procedure is initiated (comparable to the procedure in case of plagiarism, ghost writing, etc.).
• What happens if I used AI tools even though they were not permitted as aids?

  The same rules as in case of cheating apply. If the teacher or examiner suspects that you used unauthorised aids, a study law procedure is initiated. For the purpose of this procedure, observations and evidence are collected. You have the right to make a statement as well. If it turns out that you did not observe the law, an ‘X’ will be entered in your transcript of records and the exam counts as an attempt.