What is generative AI?
Generative AI means that, for example, text can be generated based on questions and statements expressed in natural language. Generative AI uses language models trained on text of varying scope and quality and produces probability-based answers according to how human language typically appears. This means that the generated responses in tools such as ChatGPT are not true or false, but probable. They are based on what is statistically the most likely next word given the prompt entered. The responses may be factually correct, but they may also be fabricated. They may also reflect stereotypes or be biased toward a certain perspective, depending on what the data in the language model looks like.
Some AI tools combine a large language model with a specific dataset. One example is Scopus AI, which is based on the content of the Scopus database. Scopus AI refers to scientific studies and can help you explore a topic, but it cannot be used for more structured searches. Even though Scopus AI is built on the content in Scopus, its AI-generated answers can be misleading, incorrect, offensive or biased. You need to make your own assessment of the responses you receive.
Always check with your teacher and/or course coordinator what is okay before using AI tools in your studies.
Scientific information searching and generative AI
When searching for scientific information, you need to be familiar with different types of scientific publications and where to find them. You also need to have knowledge of search techniques, such as how to use keywords to perform your searches and how to evaluate the material you find.
Use databases containing scientific material to search for articles. Think of generative AI as a complement to, not a replacement for, traditional scientific information searching. You can use generative AI to get tips on aspects within a subject area or suggestions for search terms to use. However, to progress in your learning, you need to use scientific material from reliable sources.
Structured information searches are characterised by their ability to be reproduced. You must be able to describe where and how you have searched to find your sources. A scientific database contains a certain type and a certain amount of curated material and provides results based on your search criteria. In an AI tool such as Scopus AI, the results you get may differ from one time to another, even if you ask the same question. The sources provided are a selection based on criteria that are not transparent.
- Read more about how to search for scientific material.
- Read more about conducting structured searches.
- Read more about different types of publications.
Generative AI, source criticism and critical thinking
Different sources of text material base their response results on different types of data. This applies to both traditional scientific databases and AI tools. For scientific databases, it is often clear which material is included in the database. For AI tools, this is often less transparent, and there is greater uncertainty in how the answers are generated. Depending on the type of AI tool, you may also need to verify that a specific source actually exists and is not AI-generated based on probability.
You always need to critically assess the responses you receive and the material you find. Are the answers reasonable? What are the sources for the claims? Do the sources exist? Is the material scientific and of good quality? AI can also be used to create content that appears credible but is false and intended to facilitate the spread of misleading information.
All data you input into an AI tool can potentially be used further, for example to train the language model. Do not enter personal data or other sensitive information. Also do not input copyrighted material.
AI tools impact the environment, for example they require large amounts of energy. Consider how you use them and what is necessary and what is not.