Fraud in Science

Alexander Berezin berezin at MCMAIL.CIS.MCMASTER.CA
Tue Jul 11 11:29:33 EST 1995


On 10 Jul 1995, William Tivol wrote:

> U27111 at uicvm.uic.edu wrote:
> 
> Dear Kathy,
> 
> : fails/refuses to recognize his/her mistakes... also, in today's
> : funding environment - mistakes are not allowed to be acknowledged].
           [...many deletions... ]
> 
> : And how many more Gallos are out there which hasn't even been
> : discovered yet?
> : ...............
> 
> 	Either too many, or zero.  I'm not betting on zero.
> 

"Either too many, or zero". Glad that you bring up this issue.
There was an interesting letter a while ago (in Nature 
magazine ?) with approximate title "FRAUD IN SCIENCE MUCH
MORE COMMON THAN WE MAY THINK" . The author's logic was 
approximately this:
  
 We, scientists, usually agree that we should test hypotheses
against available experimental data. And if there are two (or
more) CONTRADICTORY hypotheses which fit the data, they BOTH
(or all) must be considerd seriously, perhaps with equal
weight.

   In this case the "Experimental Data" are:

 "Cases of fraud do occur, but judging from press 
  reports they seem to be isolated occurences and 
  there frequently is relatively low"

At least TWO oppisite hypotheses can be drawn from this 
"Experimental Results" section:  

Hypothesis # 1
 
  "The above Exper.data INDEED mean that the fraud is
  a (relatively) rare phenomena. Overwhelming majority
  of scientists are honest reporters, don't bend or trim
  their data, don't steal (unattributedly) ideas of
  others, etc"

Hypothesis # 2

  "The above Experimental Data have a systematical error
  bias, BECAUSE the existing mechanisms of how scientific
  research is conducted are such that the DETECTION of
  fraud, plagiarism, etc. is a relatively rare happening
  in itself (too high level of "noise"). Contrary to
  what is seemingly offred by the above Data, the fraud
  is quite common, with OVERWHELMING MAJORITY OF IT 
  GOING UNDETECTED"

Both hypotheses are in full agreement (at least at this
stage) with the "Experiment" and, correspondingly, both,
at best, should be held concurrently. We can look for
further data and support arguments (direct and indirect),
for and against these opposite interpretaions.

(indirect) argument in favor of hypopesis # 2:

Most (perhaps, Lions share) of shoplifting goes undetected
and unpunished. On the basis of some (apparently, anonymous)
questionners a recent sociological article deduces that
about 10 % of all population shoplift and this figure is
not much varying for different social and income groups 
(sic !). In short, rich (and educated) shoplift as 
often as poor and uneducated.

Infering this finding to science we can
suggest 3 alternatives:

1) "shoplifting" in science significantly less than 10 %     
2) it is "about the same"  (e.g. close to 10 % )
3) it is significantly HIGHER that 10 %

My own (again indirect) argument in favor of 3) is that
it is much easier to shoplift in science that in a shopping
mall (where you, as of now, at least surrounded by hidden
cameras). Correspondingly, applying principle of the least 
resistance (cf. Fermat principle in physics), the 
inference 3) seems to be the likeliest.

Alex Berezin  
   





More information about the Bioforum mailing list