Randomly Generated Math Article Accepted By 'Open-Access' Journal 197
call -151 writes "Many years ago, a human-generated intentionally nonsense paper was accepted by the (prominent) literary culture journal Social Text. In August, a randomly-generated nonsense mathematics paper was accepted by one of the many low-tier 'open-access' research mathematics journals. The software Mathgen, which generated the accepted submission, takes as inputs author names (or those can be randomly selected also) and generates nicely TeX'd and impressive-sounding sentences which are grammatically correct but mathematically disconnected nonsense. This was reviewed by a human, (quickly, for math, in 12 days) and the reviewers' comments mention superficial problems with the submission (PDF). The references are also randomly-generated and rather hilarious. For those with concerns about submitting to lower-tier journals in an effort to promote open access, this is not a good sign!"
Re:Argument (Score:5, Interesting)
Bravo, good sir! Another milestone!
Re:Argument (Score:4, Interesting)
The obvious next step beyond randomly generated journal submissions is, of course, randomly generated Slashdot comments.
I dunno. I keep getting the feeling that literary criticism has worked that way for decades...
Re:View from the outside (Score:5, Interesting)
You are not too far from the truth.
Almost any paper presented at a conference will be intentionally opaque. If it wasn't, many people that heard the presentation would quickly write their own version and submit it to a bunch of journals hoping to claim the result as their own.
The second problem occurs because the authors have been living with the results for months working out the last few details. All the intermediate steps are obvious to them because they have been thinking about them for so long. They forget that the intermediate steps are not obvious to everyone else.
The third problem is that everyone is writing for the experts in the field. People are afraid to write the intermediate steps because of peer pressure. If they write them down, they think their peers will laugh at them and conclude their work must not be worth much.
When I was a grad student, my adviser always wanted me to read his papers. From my point of view they were a bunch of disjoint unrelated paragraphs. The few times I was able to figure out how and why he went from one paragraph to the next, I gained more insight than I did from a year of taking classes. I used to tell my peers that all the true knowledge was between the paragraphs.
who was the referee? (Score:5, Interesting)
This is massively a failure of the editor and the referee. I suspect the editor didn't look at it all and the referee did a quick superficial job. One big question is who the referee was. One typical method of finding an appropriate referee is to look in the references. However, in this case, since the references are hilariously bogus:
"[7] "Q. Hausdorff and C. W. Turing. Advanced Combinatorics. Guyanese Mathematical Society, 2001"
I don't think you are going to find a Turing or Hausdorff alive and replying to email requests to referee these days! I can't believe a mathematically literate editor would look at the references (to find a referee) and not immediately realize that this is nonsense. So I suspect the editor asked someone else who had recently submitted something to the journal to write a quick report, perhaps in the spirit of mutual back-scratching. Perhaps that referee also did not notice that this was nonsense and did not look at the references either. Or perhaps the editor did a quick review instead of sending it out- the chance that two reasonable math people, no matter how overworked with their own tasks, would not notice that this was totally bogus I would hope is small.
There is a blind spot here in our understanding of (Score:4, Interesting)
knowledge.
You cite meaningfulness and utility as two things that a randomly generated paper lacks.
Yet that is precisely what is at issue here, and what was at issue in Social Text—someone found these randomly assembled texts to be nominally meaningful, and the value of "meaningfulness" (bringing meaning to life, understanding the meaning of the universe, making the 9/11 deaths meaningful, etc.) is not zero, hence we can assume that meaningfulness is a dimension of some understandings of "utility."
Despite the intent of these kinds of papers, they appear instead to confirm at least some of the postmodernist argument: that in practice for humans, meaning and utility do not necessarily not vary either directly or inversely with enlightenment-style formal logic and or empiricist epistemology (whatever our ideals or desires), but instead that there appears to be a strong dimension of social construction involved in discerning meaning and utility, and conversely, that in many cases the things that we construct become by definition meaningful and useful in some sense as a matter of someone having constructed them, the awareness of this, and the reliance of these constructions on existing worlds of taken-for-granted meaning (language, culture, etc.)
This is not to say that "all things are equally true" or "all things can be equally true" but rather that "practical truths in social existence are never merely empirical substances" and we would do well to understand this if we want to understand/influence/improve society.
Re:Argument (Score:2, Interesting)
Obligatory XKCD? Here it is! [xkcd.com]
Re:View from the outside (Score:4, Interesting)
Some individuals may not understand the intermediate steps if they aren't intimately familiar with the field, e.g., someone new to probabilistic models may not know why you can rewrite Sethuraman's sum-based, stick-breaking construction of the two-parameter Poisson-Dirichlet/Pitman-Yor process or the one-parameter Dirichlet processes in a multinomial-based, stick-breaking form. Nevertheless, that does not necessarily mean that the context or contributions of your work won't be unknown to others.
To elaborate, I recently wrote a paper wherein I used copious amounts of differential geometry to recast a high-level machine-vision methodology in a more general, conducive fashion, then proceeded to extend and use the tools of the field to massage that scheme so that its algorithmic implementation would have a much lower computational complexity. Although the paper was sent to the top-tier computer vision/pattern analysis journal, which has been host to a few articles that make use of differential geometry, I doubt that most of the readers will care about the pages of theorems and derivations, as most are not mathematicians, and, instead, just home in on the two important, end-product equations I list, either code them up or download my code, and find that they produce the same outputs but with the new version requiring fewer calculations; further, In this case, while they may not fully grasp how I moved from one representation to the other, they can at least see that the end result is bonafide and incorporate my scheme in their work.