Position: We Should Read and Write Papers Differently in the Age of AI

Michael Eichelbeck, Tobias Ladner

Keywords: position, LLM, AI, paper writing, paper reception, scientific process, paradigm shift

TL;DR: Research papers should shift from author-centered to reader-centered formats using AI tools to provide up-to-date and personalized instead of static and potentially biased contextualization.

Abstract:

Research papers follow conventions that are widely agreed on and enforced by the scientific community.
The established paradigm is that each paper is self-contained, including contextualization of the proposed method. This static author-side contextualization creates several risks, such as incentives to promote a certain biased narrative or the outdatedness of related literature. Furthermore, readers might have to rely on external resources to assist their understanding if they have a substantially different background from the authors.
At the same time, the amount of research output is rapidly increasing in tandem with the usage of AI tools for interacting with papers.
We argue that the age of AI enables a paradigm shift in which research dissemination transitions to a more reader-centered approach, and that this shift should be approached proactively by the scientific community.

Primary Area: other topics in machine learning (i.e., none of the above)

Submission Number: 17631