Automated Research!



During the course of the Research Methodology lectures in IIT Madras, we get pointers on how to conduct a literature review and a few leads to choosing a problem to work on. Talks mention how to use the resources in the institute and on the web for scouring the research area of interest for a problem. But few people ever told us about a certain software which goes by the name SCIgen. SCIgen is an Automatic Paper Generator. Given just the names of the Authors(?) who require a paper to be published in their names, it automatically generates a paper complete with title, abstract, body and conclusions. To make this more realistic, a few graphs and equations and some authentic looking references are also thrown in.

The creators are three graduate students Dan Aguayo, Max Krohn, Jeremy Stribling of the PDOS group at MIT. They admittedly got bugged by the constant mails sent from the 'spam'ference WMSCI, so much so that they wrote a s/w to generate papers to expose such conferences. And SCIgen was born. The rest is history. Their paper was 'accepted as a non-reviewed paper' in WMSCI 2005. Buoyed by their success, they set up a website publicizing their success, and put this s/w in public domain. They also solicited funds from interested donors so that they could attend the conference (whose registration fee was quite high at $390). The site evidently became very popular (they got $2400 in 3 days), but success had its own drawbacks... the conference now rejected their paper, as one of the reviewers who got wind of this page mailed his indignation to the conference organizer. The conference fee was paid, but the organizers returned the amount saying the paper was now rejected as the author admitted it was fake. But the amount of support for their endeavour emboldened our protagonists enough to stage a fake {randomly generated} parallel conference in the same hotel where WMSCI was held. A hilarious account of the happenings there is available on their blog in the SCIgen webpage {http://pdos.csail.mit.edu/scigen/blog/}.

This phenom throws up many issues, but the most important point to emerge is that these 'spam'ferences are able to con the research community because of the enormous amount of importance put on the quantity of publications rather than the quality of research.

As side note, you can also look into an interesting piece of s/w called CS Research Topic Generator available at http://www.cs.purdue.edu/homes/dec/essay.topic.generator.html.

This is a method to generate research titles loaded with buzz words. The method is simple enough, form 3 columns, the first adjectives(viz. integrated, parallel, randomized), the second and third standard CS jargon like (mobile, wireless, algorithm, networks, neural nets etc). Now combine these three to form a phrase- one word from each column. Throw a few add ons like 'A' 'and' 'The' and viola! you have a research topic to explore. Eg: Integrated Parallel Neural Nets. More information is available in the website.

Although these appear to be research efforts worth the Ignobel, the possibility that a new area of Automated Research may evolve from these controversial beginnings cannot be discounted yet.

PS: Last line was meant to be a joke.