X

Publishers remove gibberish computer-generated research papers

Springer and IEEE will be removing more than 120 papers from their subscription services after a researcher found they were computer-generated gibberish.

Michelle Starr Science editor
Michelle Starr is CNET's science editor, and she hopes to get you as enthralled with the wonders of the universe as she is. When she's not daydreaming about flying through space, she's daydreaming about bats.
Michelle Starr
2 min read

(Credit: (Bios [bible] installation image by robotlab)

Springer and IEEE will be removing more than 120 papers from their subscription services after a researcher found they were computer-generated gibberish.

For a layperson, looking at scientific papers can be an exercise in humility. We know most of those words, and surely they make sense in some capacity, but high concept research uses, by necessity, some very complicated language.

Apparently not even the publishers of these papers are as adept as we thought at gauging their meaning, as the work of one researcher reveals. Computer scientist Cyril Labbé of Joseph Fourier University in Grenoble, France, spent two years examining published research papers, and found that computer-generated papers made it into more than 30 conferences, and over 120 have been published by academic publishing houses — over 100 by the the Institute of Electrical and Electronic Engineers (IEEE) and 16 by Springer.

The papers were generated by a piece of free software called SCIgen, developed in 2005 by scientists at MIT. SCIgen randomly generates nonsense papers, complete with graphs, diagrams and citations, and its purpose was to demonstrate how easily conferences accept meaningless submissions.

Labbé, who has built a website that allows users to check whether a paper was generated using the program, said he did not know why the papers had been submitted. Most of the conferences took place in China, and many of the papers named real authors, some of whom may or may not have known their names were being used in this way. One author claimed he had created the paper to test a conference.

Both publishing houses have pulled the papers in question, although some issues remain. Ruth Springer, UK head of communications for Springer, is running into issues trying to contact the authors, and noted that the conferences in question were, in fact, peer reviewed, which casts a bad light on the current processes. IEEE, however, declined to comment.

Labbé's research was published in a paper titled "Duplicate and fake publications in the scientific literature: how many SCIgen papers in computer science?" Meta.

Via www.nature.com