In a stunt reminiscent of Alan Sokal’s article published in the journal Social Text, three MIT grad students in computer science were invited to present at an academic conference on the basis of a paper randomly generated by a program they wrote. Reuters reports (via CNN.com):
[Co-author Jeremy] Stribling said the trio targeted WMSCI because it is notorious within the field of computer science for sending copious e-mails that solicit admissions to the conference.
The idea of a fake submission was to counter “fake conferences…which exist only to make money,” explained Stribling and his cohorts’ website, “SCIgen – An Automatic CS Paper Generator.”
“Our aim is to maximize amusement, rather than coherence,” it said. The website allows users to “Generate a Random Paper” themselves, with fields for inserting “optional author names.”
Their program is easy to use. After a few tries I came up with a paper suitable for presentation:
A Methodology for the Emulation of B-Trees
Michael Giberson, Kevin Bacon and Paul Erdos
The theory method to IPv4 is defined not only by the analysis of link-level acknowledgements, but also by the significant need for rasterization. Given the current status of amphibious models, end-users predictably desire the improvement of evolutionary programming. In this paper, we use wearable models to show that the little-known Bayesian algorithm for the simulation of DNS by F. Ito et al. is impossible.
The computer-generated paper continues with a table of contents, six sections of text, five figures, and a list of 21 references. The paper even cites my “previous paper”: “Giberson, M. Decoupling B-Trees from the lookaside buffer in consistent hashing. Tech. Rep. 565-81, UCSD, Dec. 1995.”
(I hope having some famous co-authors will help me land a primo spot on some conference agenda.)
Tip o’ the pen to Dan Houser for alerting me to the story.