AI in Science: The 42 problem in AI-assisted research
- ITE Lab, Institute of Biomedical Ethics and History of Medicine, University of Zurich, Switzerland
- Institute for Data Science and Artificial Intelligence, Boğaziçi University, Turkiye
Figures
Figure 1
The value of a synthesis is not monotonic in its compression.
(A) As the length L of a synthesis falls, recovery R(L) declines from saturation, while the reading cost C(L) falls steadily, so that value initially rises. This is the case for automated synthesis, and within this range it is correct. Usability U(L) holds near one until L crosses the threshold L*, below which the text no longer carries the conditions that make its claims valid, and V(L)=R(L)·U(L)/C(L) collapses. (B) The position of L* is set by the question and the reader, not by the text: the same synthesis sits above the cliff for a screening decision and below it for a clinical guideline. Links do not raise the curve; they displace context rather than destroying it, moving L* to the left and extending the range over which compression remains safe.
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
AI in Science: The 42 problem in AI-assisted research
eLife 15:e113137.
https://doi.org/10.7554/eLife.113137
