Your Name and Title:
Jesper Solheim Johansen, Head of User Research
Library, School, or Organization Name:
Keenious
Area of the World from Which You Will Present:
Norway
Language in Which You Will Present:
English
Target Audience(s):
Anyone interested in the effects of generetaive AI on higher education
Short Session Description (one line):
This session investigates the "citing without sighting" paradox
Full Session Description (as long as you would like):
For the last couple of years, there has been a lot of talk about generative AI. Many in libraries and learning centers were probably left with the impression that it was mostly hype, as the everyday life of the library didn't seem to change all that much. However, that has changed recently.
In 2025, several studies showed a significant increase in the number of references included in student assignments. At the same time, many libraries are seeing the opposite trend: an overall decrease in the download of full-text journal articles from their collections. These are very contradictory numbers. If students are citing more, why are they downloading less?
It is reasonable to assume that this trend comes from the common use of AI in information gathering. Instead of reading the actual papers, students are now using generative AI tools as a "layer" on top of the information. The language model finds the paper and rephrases it so that students can paraphrase directly into their own text. Often without ever touching the original source.
The optimists say this is simply the new information landscape. They argue that engagement via an AI layer is as good, or even better, than the old way of working. On the other hand, critics argue that it is worrying when students lose their touchpoint with the original source. Receiving information only through the lens of an AI introduces biases and a great risk of "cognitive offloading", where convenience is prioritized over actual learning and comprehension.
In this talk, we will take a closer look at these recent trends in citations and downloads. We will weigh the different arguments and discuss whether generative AI can be designed in a way that helps researchers without removing the essential engagement with the original material.
Replies