What do you have to think about when using an AI-driven text generator? How can artificial intelligence enrich your learning process – and what is not recommended?
At this weeks “Brains & Breakfast” Martin K. Steinsmo, academic librarian at NMBU, presented how AI tools can be useful for students who want to brainstorm ideas, learn new things, or improve their writing skills. However, AI tools also pose some challenges and risks that you should be aware of before using them in your academic work. Here are the main take away points from the presentation:
Check what you are allowed to do
At NMBU there are no unified guidelines for using AI in student work. It is up to each faculty and course teacher to determine what is allowed in each course. You should always check with your faculty or teacher which rules apply for using AI tools in your field of study and your academic work. The tips in this article are only reflections about best practice if use of AI are allowed in the courses you take.
- Make yourself familiar with NMBUs permitted aids.
- Guidelines for use of artificial intelligence at REALTEK
Know the bias
The AI tools all have biases. You should always be aware of the biases in AI models, and critically review the output to detect biased content. When you are prompting and providing the tools with context, make sure to use neutral language and avoid asking leading questions.
Check the facts
Never blindly trust the generated content, and always double-check the facts with academic sources. When you cite, cite the sources you have used to verify the content. Do not refer to AI generated content as facts in your assignments.
Work in an ethical way
You should use AI tools in a responsible and ethical way, and not as a shortcut or a substitute for you own work. You should never copy and paste generated content directly into your assignment or use these tools to completely rewrite your text. The AI tools can be used as a starting point or a finishing touch. Your own analysis, evaluation, and synthesis is the most important part of your text.
Be honest and transparent
You should always be transparent and honest about how you used AI tools, and describe which tools you used, what specific prompts you gave them, and how you used the generated content in your assignment.
GDPR
AI tools are mostly provided by US-based companies. Norway has stricter privacy laws than the US, as we have to follow both national and European regulations on this issue. Not all tools can comply with these strict rules. ChatGPT, the most popular generative AI model, is one of the services that does not have enough safeguards for storing and using personal data. The University has therefore obtained a Norwegian developed GPT service, called Sikt KI chat.
SIKT KI-chat – the recommended tool
This software is based on the GPT-4 technology and can offer similar features as ChatGPT. The software also follows Norwegian privacy laws. Every student and employee at NMBU can access can access this tool from this website: https://ki-chat.sikt.no/en. You can chat with the model in English and Norwegian and keep chats for up to 30 days.
To sum up; Test – have fun – fact check– and remember critical thinking and ethical awareness!
Facts:
The draft for this article was generated by Microsoft Copilot, based on the presentation by Martin Steinsmo. The text has been rewritten and checked by NMBUs librarians.
The presentation focused especially on generative AI, which is a term that is used to describe «[…] models that can produce new, previously unseen information based on the data they were trained on» (Garcia-Peñalvo, F.J. & A. Vazquez-Ingelmo, 2023).
For example, generative AI can create text, images, or sounds that mimic human creativity and intelligence. The most famous Generative AI models are ChatGPT (text) and Dall-E (pictures).
Reference:
Garcia-Peñalvo, F.J. & A. Vazquez-Ingelmo (2023). «What Do We Mean by GenAI? A Systematic Mapping of The Evolution, Trends, and Techniques Involved in Generative AI”. International Journal of Interactive Multimedia and Artificial Intelligence, 8 (2023), 7-16.