Author: Theo Pavlidis
Position: Distinguished Professor Emeritus
Institution: Stony Brook University
E-mail: t.pavlidis@ieee.org
Submitted Date: October 20, 2008
Published Date: October 21, 2008
This comment was originally posted as a “Reader Response” on the publication date indicated above. All Reader Responses are now available as comments.

I believe that the article by Young, Ioannidis, and Al-Ubaydli touches important issues that extend beyond the bio-medical sciences. Early in my career I collaborated with biologists and, amusingly, my only publication in "Nature" (v. 259 in 1976, pp. 343-344) is as a co-author of a letter to the editor pointing out the fallacy of certain arguments in a published paper. That paper presented a "surprising" result that has long been consigned to oblivion. Thus the Young et al paper "hit a spot"!

However, I am not writing to relate only a personal story. Most of my career has been spent in Computer Science and Engineering and I recognize a similar pathology in that field as that described in the paper. The tendency in Engineering/Technology research is to prefer papers with general solutions. That would have been fine if it were not for the fact that many of these general solutions are obtained on the basis of unrealistic assumptions or are insufficiently documented. There is no follow up to see if any of the results are used in practice - the counter part of the medical efficacy of biomedical results. As a result there is a gap between academic technology and industrial technology, to the detriment of both. Researchers who do both academic and industrial research are a rarity.

I have held several editorial positions, including editor-in-chief of the IEEE Transactions on Pattern Analysis and Machine Intelligence (1982-86), considered the top journal in its field, and member of the editorial board of the IEEE Proceedings (1988-1997). While I was keenly aware of the problem of inflated results, the social dynamics of the publication process were such that I could do little but deal only with the most extreme cases. Therefore I doubt that much can be done to reform the publication system under its current form.

I think that the only long-term solution lies in open publication on the web and the reliance on search engines such as Google to identify articles of interest, a point made also by Professors Gordon and Poulin. Tools such as Google Analytic could be used both by academic authorities and funding agencies to track the readership of a publication. (While Google Analytic is intended for analysis of the traffic by the owners of web sites, it can be easily adapted for the stated goal.) While extravagant claims may attract initially a large number of readers they will prove eventually embarrassing to the authors, thus the hope is that authors will not make them. Now are necessary for a publication in a prestigious journal that is a goal in itself.

Web publication (in combination with search engines) would implement several of the solutions listed in Box 1, particularly 2, 5, and 7.

I should add that since my retirement from both my academic and industrial positions I have published only on my personal web site and I have been pleasantly surprised by the number of readers (several thousand in some cases) and by their feedback. This has happened not only with publications in my area of expertise where I may have had an advantage in attracting readers but also in publications in entirely new areas that retirement gave me the luxury to pursue.