by Paula Dorfer
The workshop opened with a check-in round in which fellows shared their current approach towards using AI for their research, the concerns that still accompany their use, and the benefits they hoped AI could offer them moving forward. While most participants already rely on AI for organizational tasks or translation, a number of open questions remained: for instance, around the protection of original ideas, ethical boundaries in research, the risk of losing skills or never attaining them in the first place, and the right way to disclose AI usage in academic work.
Building on these reflections, Dr. Macho guided the group through the legal and ethical issues surrounding AI in academic research, addressing copyright law, data protection, and the principles of good scientific practice. A central message was that researchers themselves must remain the initial and final authority over every use of AI. The fellows discussed in detail which forms of AI assistance can be considered ethically unproblematic, and which uses become questionable.
A further focus of the day was data protection, particularly the protection of confidential data and original ideas. Fellows learned to distinguish between private and institutional AI access, worked through the "data protection traffic light" for classifying content as safe, sensitive or confidential, and applied a concrete decision pathway to examples from their own research. Two practical settings were highlighted as immediate, actionable steps every researcher can take: disabling the training of the model with one’s input and using temporary or incognito chats.
The workshop also examined how generative AI actually works, illustrating word-by-word text prediction and explaining why AI can produce convincing yet inaccurate or hallucinated answers. Participants explored three key limitations of AI-generated support in academia: its currency and timeliness, its accessibility to content (especially in regard to paywall databases), and its research depth.
During the final part of the day, Dr. Macho introduced the three building blocks of prompting for academics and helped fellows to develop a “priming abstract”, a short summary of their research project to be added to any prompt in the future.
The session closed with a discussion of three frequent risks in AI use: agreeableness with the user, premature answers with the risk of not having understood the prompt, and the lack of transparency and reliability of the path of its own reasoning.
Through a mix of theoretical input, group discussion, and hands-on exercises, the workshop gave Ars Iuris fellows a well-rounded and practical foundation for using AI responsibly in their research, leaving participants with concrete strategies they can use in their own work.
