Arqus "AI in teaching & learning: International perspectives and best practices” - Workshop #4

Описание: Artificial intelligence is changing university teaching – but how can it be used effectively? In this series of online workshops, experts from Arqus universities share their experiences and challenges in using AI in teaching. The fourth session of the series took place on 15 January at 14:00 CET on Zoom.

The fourth session featured four talks on four different topics:

Behind the scenes of genAI: Unmasking bias, hallucination, and AI slop in language learning, by Elke Höfler (University of Graz)

Artificial intelligence has become a constant companion in language learning – from automated translations to AI-generated feedback. Yet, with this convenience comes new responsibility.

In this talk, the speaker explores how language teacher education can foster critical awareness of AI, presenting three examples from language teacher education that invite future teachers to reflect critically on the promises and pitfalls of AI.


Custom GPT for formative feedback in undergraduate physics lab reports, by Arin Mizouri (Durham University).

This talk focuses on a study at Durham Physics that developed a prompt-engineered, course-aligned GPT to provide formative feedback on first-year lab reports.

The speaker summarises the initial student survey, highlighting low confidence in writing, show how the model was constrained with rubrics and exemplars, and present evaluation data from 15 students on its usefulness and accuracy.


From manuscript to machine partner: AI as a co-author in graduate scholarship, by Alexander Godulla (Leipzig University)

This talk explores how generative AI reshapes the production of academic books and conferences at the Master’s level.

The speaker examines AI as a co-author, research assistant, and editorial tool, showing how it can support literature work, argument development, design of scholarly formats, and the creation of high-quality academic outputs.

Making chemistry machine-readable and creating datasets for improved GenAI chatbot performance in undergrad organic chemistry, by Sebastian Tassoti (University of Graz).

For higher chemistry, it is not surprising that GenAI chatbots do not have a lot of training data so it becomes crucial how you prompt the bot.

In this talk, the speaker presents evidence and the design of an intervention done with 45 students in Graz and New York City that focuses on how to make organic chemistry machine-readable and how to write a prompt that trains an LLM to improve performance and use it to predict reactions with higher accuracy.

#artificialintelligence #ai #teaching #arqusalliance #university #genai