On Sun, Aug 22, 2010 at 11:47 PM, Andrew Odlyzko <odlyzko_at_umn.edu> wrote:
> There is perhaps an even more important reason not to worry about this effect:
>
> If university bean-counters do start using download statistics in a serious
> way, those statistics will quickly become distorted to the point of being
> meaningless. We already have extensive evidence that impact factors of
> journals and citation counts are being systematically distorted. It
> would be far, far easier to distort download counts. "Mike, if you will
> have each of your students take a look at one of these half a dozen of
> my papers, I will assign your favorite papers to my classes as required
> reading, ..."
Andrew is absolutely right. All metrics can be abused. Online they can
be abused even more easily. Friends can be asked to download, or
software can be written to do it automatically.
But this is, after all, academia, and not the trade world of spamware
and fraud. So abuses have bigger costs than just the time and money
invested in the scam.
Academics are answerable, with their careers and reputations. Authors
caught out in metric abuse can be named and shamed, and I would say
the potential risks vastly outweigh the potential benefits.
For although the online medium does increase -- greatly -- the
potential for abuse (of all kinds: inflating impact metrics,
plagiarism, fraud) it also increases -- even more greatly -- the
potential for detecting and exposing the abuse.
And there is strength in numbers. Research impact is not a
1-dimensional scalar. It's a vector with many interdependent
components, not all of them equally easy to abuse singly; and the
overall pattern, with its interactions and interdependencies, is
virtually impossible to abuse.
Just as self-citations, co-author citations and citation circles are
readily detectable and identifiable (and hence separable from
independent citations) so download manipulations will become
increasingly detectable: For example, you can fake higher download
counts with bogus hits, but you cannot fake the higher citation counts
that higher downloads normally generate after a certain latency. A
download burst not followed by a citation burst is a detectable
anomaly. So are multiple downloads to the same site or to nonacademic
(or faked academic) sites.
Brody, T., Harnad, S. and Carr, L. (2006) Earlier Web Usage Statistics
as Predictors of Later Citation Impact. Journal of the American
Association for Information Science and Technology (JASIST) 57(8) pp.
1060-1072.
http://eprints.ecs.soton.ac.uk/10713/
The reason most online spams and scams flourish is that, if caught,
their consequences are no more than a lock-out by one service
provider. The consequences of academic monkey business represent a
rather more sobering deterrent.
Stevan Harnad
> Stevan Harnad <amsciforum_at_GMAIL.COM> wrote:
>
> > On Sun, Aug 22, 2010 at 5:59 PM, Michael Smith
> > <Michael.E.Smith.2_at_asu.edu> wrote:
> >
> > > When I was pitching self-archiving to some colleagues last week, two of them
> > > mentioned the following argument AGAINST self-archiving. University
> > > bean-counters have started using the number of times articles are downloaded
> > > (from publishers sites, I guess) as a measure of faculty productivity or
> > > impact. If one self-archives, then people will be less likely to download
> > > from the publishers site, thereby lowering one’s download score. I can think
> > > of various reasons why this is NOT a good reason to avoid self-archiving,
> > > but I wonder if there are any data on this, or if any bibliometric
> > > researchers have addressed this topic explicitly.
> >
> > Here are just a few reasons (each one of them a no-brainer):
> >
> > (1) More accessibility does not decrease total downloads, it increases them.
> >
> > (2) OA self-archiving, while increasing total downloads, may shift
> > some of the download traffic from the publisher's website to the
> > institutional repository.
> >
> > (3) Download counts from the institutional repository can be added to
> > download counts from the publisher's websites.
> >
> > (4) Open Access self-archiving also increases citations -- another,
> > more venerable target of the bean-counters.
> >
> > (5) Increased downloads lead to increased citations.
> >
> > How many more reasons do the bean-counters need, to mandate OA self-archiving?
> >
> > > Michael E. Smith, Professor
> > > School of Human Evolution & Social Change
> > > Arizona State University
> > > www.public.asu.edu/~mesmith9
Received on Mon Aug 23 2010 - 17:15:30 BST