Hot and trending as this topic may well be, the use of (Gen)AI in research is here to stay. And it’s use also presents some challenges from a research integrity perspective. What are the (dis)advantages and how can we integrate them into good research practice? Who is responsible for the correct use of (Gen)AI and how can we achieve the correct use of the tools amongst researchers?  Walter Daelemans, Full Professor of Computational Linguistics at the University of Antwerp, Geert-Jan Bex, Full Professor at the Faculty of Science, Hasselt University and Vincent Ginis, Associate Professor at the Data Analytics Lab at Vrije Universiteit Brussel discuss the main issues from their own research expertise.  

In this episode, we cover:

• What is (Gen)AI & how does it work?
• How can (Gen)AI be used in an academic (research) practice?
• What is the impact of the use of (Gen)AI on our time and efficiency?
• Ways to estimate the extent to which (Gen)AI is used
• What is responsible use of (Gen)AI?
• What do you need to know before starting to use (Gen)AI?
• What disadvantages should you take into account?
• What are the main ethical issues when using (Gen)AI in research?
• How to deal with hallucinations in tools?
• How to reference the use of (Gen)AI?
• Who is responsible when using (Gen)AI?

---------------------------------------------------------------

This podcast series complements the online training tool 'Mind the GAP, training on Good Academic Practices'. Mind the GAP is an English-language training tool for all researchers and those involved in research, from PhD students to more experienced researchers, to teachers and policy makers.  

If you are affiliated with a Flemish university you can find the tool on your institution’s educational platform: 

Ghent University: Ufora 

KU Leuven: Toledo 

University of Antwerp: Blackboard 

Hasselt University: Blackboard  

Vrije Universiteit Brussels: Canvas  

Not part of the above institutions? Go to https://mindthegap.vlir.be/ and follow the international version of the tool (condensed version).   

The Mind the GAP Podcast was jointly developed by VLIR (Flemish Interuniversity Council – Filip Colson) and the five Flemish universities (Ghent University – Stefanie Van der Burght; KU Leuven – Wouter Vandevelde; University of Antwerp – Marianne De Voecht; Hasselt University – Stephanie Ruysschaert; Vrije Universiteit Brussel – Klara Swalus) and was financed by the Flemish government. It was produced by podcast agency De Praeters and hosted by Elisa Nelissen (KU Leuven).  

Connect with Us: https://mindthegap.vlir.be/

------------------------------------------------------------

Resources Mentioned 

• ALLEA (2023). The European Code of Conduct for Research Integrity: Revised Edition 2023. Berlin. DOI 10.26356/ECOC       
• Chat GPT  
• DALL-E 
• Github 
• Copilot 
• Consensus 
• Roth, E. (2024). Google explains Gemini’s ‘embarrassing’ AI pictures of diverse Nazis. The Verge, 23 February. Retrieved July 9 from Google explains Gemini’s ‘embarrassing’ AI pictures of diverse Nazis - The Verge  

Key Takeaways

• Researchers remain responsible for the use of (Gen)AI and its outputs (full accountability). Institutions have an important role to play in building the right infrastructure.
• You need to become an expert in a (Gen)AI tool to use it as an assistant (you must know your field before you can use it in your field)
• Avoid copy/paste in and out of any (Gen)AI tool. Think of it as a conversation partner, get input and work on it yourself.
• Scrutinize everything that comes out with domain knowledge.
• Garbage in = garbage out.
• Tools and models are constantly evolving, both the possibilities and concrete usability need to be monitored and fine-tuned constantly.
• Hallucinations are not necessarily a bad thing, it shows the creative possibilities of a system. It is up to the researcher to make sense of it in the context of use.
• Referencing (Gen)AI makes little sense as in the future tools will be built in all kinds of applications which will make them even more difficult (or impossible) to be distinguished.