NRLgate -
Plagiarism by Peer Reviewers
Sections 7 thru 7.8
This page is part of the NRLgate Web site presenting evidence of
plagiarism among scientific peer reviewers involving 9 different peer review
documents of 4 different journal and conference papers in the fields of
evolutionary computation and machine learning.
This page contains sections 7 through 7.8 of "Indications that there
are only 2 or 3 (as opposed to 9) different plagiarizing reviewers among
the peer reviewers at the Machine Learning Conference (MLC), the editors
and members of editorial board of the Evolutionary Computation journal
(ECJ), and the Tools for Artificial Intelligence conference (TAI)."
Go to top of NRlgate Home Page
Go to Abbreviated Table of Contents
Go to Detailed Table of Contents
Go to Complaint Letter
to the Evolutionary Computation journal
Go to top of Previous Page
Go to top of Next Page
7. Indications that there are only 3 (as opposed to 9) different plagiarizing
reviewers among the peer reviewers at the Machine Learning Conference (MLC),
the editors and members of editorial board of the Evolutionary Computation
journal (ECJ), and the Tools for Artificial Intelligence conference (TAI)
Most people (including myself) believe that wrongdoing is relatively rare
even though we know that wrongdoing occurs in every aspect of human activity.
Although it is conceivable that reviewers A, B, X, Y, #1, #2, #3, T2, and
T3 of the 4 different conference and journal paper involved here are 9 different
people, the principle of Occam's Razor suggests that the simplest explanation
for a situation warrants some extra attention.
The existence of 4 disjoint groups of plagiarizing reviewers would not make
the offense of plagiarism any less serious, it would just mean that there
are a surprisingly large number of disjoint groups of plagiarizing reviewers
within the extraordinarily small pool of people in the fields of genetic
algorithms and machine learning.
Documents A, B, X, Y, #1, #2, #3, T2, and T3 establish, on their face, that
serious scientific misconduct involving collusion and plagiarism has occurred.
Plagiarism among peer reviewers is an offense that goes to the heart of
the integrity of the scientific peer review process. Someone created these
plagiarized documents. Someone committed these offenses. Up to 9 different
persons violated the trust reposed in them by the Evolutionary Computation
journal, the Machine Learning Conference, and the Tools for Artificial Intelligence
conference. It is conceivable that reviewers A, B, X, Y, #1, #2, #3, T2,
and T3 of the 4 different conference and journal paper involved here are
9 different people. The 4 disjoint groups of plagiarizing reviewers would
be
- reviewers A and B for my Machine Learning Conference paper on empirical
discovery,
- reviewers X and Y for my Machine Learning Conference paper on optimal
control strategies,
- reviewers #1, #2, and #3 for my Evolutionary Computation journal
paper on electric circuit design, and
- reviewers T2 and T3 for my paper on pursuer-evader games submitted to
the Tools for Artificial Intelligence conference (TAI).
I believe that a definitive identification of the wrongdoers involved in
creating these plagiarized reviews should be done by an impartial person
who is experience and trained in reaching findings of fact and making judgments
based on the evidence. Specifically, I advocate a complaint resolution procedure
involving a retired federal judge acting under the auspices of the American
Arbitration Association for the task of making a definitive determination
of the truth. No final judgment or opinion should be formed at this time
on any of the matters herein. Instead, the truth concerning all of these
matters herein should be definititvely determined in a thorough and impartial
investigation and factual determination made under the proposed arbitration
procedure by a retired federal judge.
However, the reader may wish to make some of his own preliminary conclusions
as to who created the 9 peer reviews for these 4 different conference and
journal papers. The factors below may enable the reader to reach his own
preliminary conclusion concerning this matter. The discussion below falls
into two main categories.
The information in the previous
section (6 thru 6.3) may cause the reader to reach a preliminary conclusion
for himself that the small overlap of people in the pools of reviewers for
the artificial intelligence conference, the Machine Learning Conference,
and the Evolutionary Computation journal along with the high level
of familiarity with evolutionary computation exhibited by reviewers A, B,
X, Y, #1, #2, #3, T2, and T3 may suggest the identities of the plagiarizing
reviewers.
The information provided may cause the reader to reach a preliminary conclusion
for himself that reviews A, X, #2, and T2 of the 4 papers are inter-linked
by so many similarities that they were probably written by the same person.
Similarly, for reviews B, Y, #1, and T1 and for reviewers #3 and T3. That
is,
- reviewers A, X, #2, and T2 are the same person
- reviewers B, Y, #1, and T1 are the same person
- reviewers #3 and T3 are the same person
The clickable diagram below presents plagiarism as horizontal dashed lines
(with added arrows shown when the documents indicate the time sequence of
the plagiarism) and presents likely common authorship as vertical lines
for 9 different reviews of 4 different papers at the Machine Learning Conference
(MLC), the Evolutionary Computation journal (ECJ), and the Tools
for Artificial Intelligence (TAI) conference.
ONLY 2 COMMON
REVIEWERS FOR
MLC, ECJ, TAI
.................THIRD REVIEWER
...B..<---..A
...Y..<---..X
...#1.--->..#2..---..#3
...T1.......T2..---...T3
Of course, if multiple papers turn out to have been reviewed the same person,
there is no plagiarism within that group of papers. Such reviews would be
expected to resemble one another. It is bad practice, but not a crime, for
a single reviewer to excessively review a single author's work.
On the other hand, if there is no common authorship among the 9 reviewers,
then there must have been plagiarism between 4 separate groups of plagiarizing
reviewers.
7.1. Reviews A, X, #2, the signed non-anonymous review to the MIT Press
of my first book, and the North American Associate Editor's transmittal
letter employ the word "limitation" with a frequency of 1-in-530;
the 316 GP-96 peer reviews use it with a frequency of 1-in-12,822; and the
word doesn't appear at all in reviews B, Y, #1, #3, T1, T2, and T3
Everyone has certain words that he habitually overuses and underuses.
Referring to the computer file containing the 64,109 words of the 316 paper
review forms from the 86 peer reviewers of the genetic programming papers
at the Genetic Programming 1996 Conference, we found only 5 occurrences
of the word "limitation" (1-in-12,821 odds).
Here's the frequency of occurrence of the word "limitation" in
various review documents:
- Reviewer B - 0 times
- Reviewer Y - 0 times
- Reviewer #1 - 0 times
- Reviewer #3 - 0 times
- Reviewer T1 - 0 times
- Reviewer T2 - 0 times
- Reviewer T3 - 0 times
- 316 reviews by 86 GP-96 peer reviewers - 5 times (1-in-12,821 odds)
Now let's look at the occurrence of the word "limitation" for
some other peer review documents.
Reviewer A says, in his 218-word review, says
- The paper completely lacks any discussion of limitations
of the method.
Reviewer X says in his 501-word review, says
- There is no discussion of the limitations of the method
Reviewer #2 says, in his 1,488-word review, says
- The limitations of the method and topics for further research
need to
- be addressed.
John Grefenstette in his 246-word transmittal letter (to which reviews #1,
#2, and #3 were attached), says
- The limitations of the method and topics for further
research need to
- be addressed.
John Grefenstette, in his 1,787-word signed, non-anonymous review to the
MIT Press of my first book, says
- The limitation of the maximum permissible size of offspring
(p. A-106)
- is a significant aspect of the approach
John Grefenstette later says,
- the author instead
- dismisses the limitation of offspring size as a mere inconvenience
John Grefenstette later says,
- The size limitation also belies the
- author's repeated assertion that his approach makes no assumption
John Grefenstette later says,
- Another limitation that is lightly brushed aside is the selection
of
- cases
Thus, reviewers A, X, #2, and these two signed documents usesthe word "limitation"
a total of 8 times in the space of a 4,240 words (1-in-530 odds).
By the way, the reader may notice that the word "limitation" appears
in question 4 of the Evolutionary Computation journal's standard
paper review form:
- DISCUSSION. Does the paper discuss relevant earlier work, noting
- similarities, differences and progress? Does it discuss the limitations
- of the approach along with its advantages? Does it consider the
- implications of the work and outline directions for future work?
(Emphasis added).
One might think that the appearance of a word in the question on a paper
review form might create a tendency of a reviewer to use that word in his
review. However, as can be seen above, ECJ reviewers #1 and #3 did not use
the word at all.
By the way, it should be noted that the paper review form for the Evolutionary
Computation journal was originally created by the editors of the journal
at the time of its inception. Thus, the paper review form itself represents
yet another appearance of this particular word.
Of course, the overuse of a particular word does not alone establish the
identity of the author of any particular document.
7.2. Reviewers B, Y, and #1 consistently misspell "LISP"
The following observation is consistent with the equivalence of reviewers
B, Y, and #1 (and T1).
"LISP" (an acronym) was misspelled as "Lisp" consistently
9 times out of 9 by reviewers B, Y, and #1.
- 3 out of 3 times by MLC reviewer B
- 1 out of 1 times by MLC reviewer Y
- 5 out of 5 times by ECJ reviewer #1
(Reviewer T1 of the TAI paper did not use the word at all).
It should be noted that Kenneth DeJong has, from time to time, published
a series of survey articles of the field of evolutionary computation. These
lengthy articles typically contain a few sentences about genetic programming.
For example, in his 11-page paper at the 1996 EvCA conference entitled
- Evolutionary Computation: Recent Developments and Open Issues
In this 11-page paper, Kenneth DeJong (1996) summarizes all the activity
in the field of genetic programming (including all 600 papers published
by about 150 different authors) in a single paragraph (which, at the same
time, purports to explain why genetic programming cannot possibly work):
- Historically, much of the EA work has involved the evolution of
fairly simple structures [that] could be represented in phenotypic form
or be easily mapped onto genotypic representations. However, as we attempt
to evolve increasingly more complex structures (e.g., Lisp code (Koza,
1992) or neural networks (de Garis, 1990)), it becomes increasingly
difficult to define forms of mutation and recombination which are capable
of producing structurally sound and interesting new individuals. If we look
to nature for inspiration, we don't see many evolutionary operators at the
phenotype level (e.g., swapping arms and legs!). Rather, changes occur at
the genotype level and the effects of these changes instantiated via growth
and maturation. If we hope to evolve such complexity, we may need
to adopt more universal encodings coupled with a process of morphogenesis
... "
(Emphasis added).
Authors carefully check their papers prior to publication, so a spelling
error appearing in a published paper probably indicates that the author
thinks the word is spelled that way. The EvCA conference proceedings (like
those of many conferences, including the ICGA conference mentioned below)
are printed using camera-ready originals provided directly by the
author. A spell checker would not ordinarily catch an error of this kind
involving a technical acronym such as "LISP." Thus, there seems
to be little doubt that Kenneth DeJong thinks that "LISP" is spelled
"Lisp."
This is not the only published paper where we find DeJong misspelling "LISP"
as "Lisp" in a published paper. For example, in their ICGA-93
conference paper entitled,
- On the State of Evolutionary Computation.
Kenneth DeJong and William Spears mentioned,
- ... Lisp code (Koza, 1992) ... (page 621)
(Emphasis added).
Thus, all 9 occurrences of this word are misspelled in MLC reviews
B and Y and ECJ review #1. And, this word is similarly misspelled in these
published EvCA and ICGA papers.
In contrast, all 4 of his references to the LISP programming language
by ECJ reviewer #2 are spelled correctly. Similarly, the 1 reference to
LISP by MLC reviewer A is spelled correctly. Similarly, the 2 references
to LISP in the signed, non-anonymous review to the MIT Press of my first
book written by John Grefenstette are spelled correctly. However, MLC reviewer
X and TAI reviewer T2 each used this word once and misspelled it. Thus,
7 out of 9 spellings are correct among this last group of 5 documents.
Of course, a repetitively-made spelling error does not alone establish the
identity of the writer of reviews B, Y, and #1. It should be noted that
I am not claiming that an individual writer can be picked out of a large
group by virtue of a single idiosyncratic spelling error. However, when
it gets down to grouping documents for the purpose of associating them with
1 of 2 possible persons, such personal trademarks can be useful. The following
sections contain additional such trademarks.
7.3. Reviews B, Y, #1, and T1 do not contain any typographical or spelling
errors (except for "Lisp"), whereas reviews A, X, #2, #3, and
the signed, non-anonymous review to the MIT Press of my first book all contain
numerous such errors
Some people are very accurate typists (or use a spell checker) when they
write a peer review. Others are not. For example, I usually do not run my
spell checker when I write a peer review.
There are no spelling or grammatical errors in reviews B, Y, #1, or T1.
Reviewer A says,
- This paper reorts.
- alternative search technqiue
- The methodology adopted prevents a clear assessment of how much over
advance this approach represents.
- (Reviewer A apparently meant to type "of advance over" instead
of "over advance")
- (All spelling errors in original throughout this section).
- (Emphasis added throughout this section).
Reviewer X says,
- The papers presents one example
- The technical soudness of the paper
Reviewer #2 says,
- automatic function defintion
- Woudln't
Reviewer #3 says,
- wikch
- excrutiating
- too lengthly
The signed, non-anonymous review to the MIT Press of my first book written
by John Grefenstette says,
- mathematic formulas
Of course, being a careful typist (or using a spell checker) does not alone
establish the identity of the writer of reviews.
7.4. Reviewers A, X, and T2 use the particular infrequently used word
"judge"
Everybody has their favorite words.
I was surprised to learn how rarely the seemingly ordinary word "judge"
is used by contemporary scientific peer reviewers. When I examined the computer
file containing the 64,109 words of the 316 paper review forms from the
86 peer reviewers of the genetic programming papers at the Genetic Programming
1996 Conference, there were only 3 occurrences of this particular word in
64,109 words (1 in 21,379 odds).
Reviewer X of my MLC paper on optimal control strategies used this word
twice in his 501-word review:
- The technical soudness of the paper is extremely hard to judge.
- ...
- The data provided is insufficient to judge the merits of this
approach.
(Spelling of "soudness" in original)
(Emphasis added throughout this secton).
.
Reviewer T2 of my TAI paper on pursuer-evader games said,
- Good solutions are achieved so rapidly that is hard to judge
the difficulty of the problems (a comparison with random search
or other techniques would be helpful).
Reviewer A of my MLC paper on empirical discovery said,
- In order to judge, it would be necessary to see the results
compared against an alternative search technqiue, perhaps even random
search.
(Spelling error in "technqiue" in original)
(Emphasis added).
Here's the frequency of occurrence of the word "judge" in various
review documents:
- Reviewer B - 0 times
- Reviewer Y - 0 times
- Reviewer #1 - 0 times
- Reviewer #2 - 0 times
- Reviewer #3 - 0 times
- Reviewer T1 - 0 times
- Reviewer T3 - 0 times
- Reviewer T4 - 0 times
- 316 reviews by 86 GP-96 peer reviewers - 3 times (1-in-21,379 odds)
Of course, the overuse of a particular word does not alone establish the
identity of the plagiarizers involved here. I am not claiming that an individual
writer can be picked out of a large group by virtue of frequent use of a
single infrequently used word. However, when it gets down to grouping documents
for the purpose of associating them with 1 of 2 possible persons, such personal
trademarks can be useful.
7.5. Review X and the signed non-anonymous review to the MIT Press of
my first book used the unusual self-referential style "This reader"
There are many ways to express the same thought; however, some figures of
speech are rarer than others. Some are so unusual that they rarely, if ever,
appear in contemporary writing by scientific peer reviewers. The self-referential
style "this reader" is one such phrase.
We again make reference to the computer file containing the 64,109 words
of the 316 paper review forms from the 86 peer reviewers of the genetic
programming papers at the Genetic Programming 1996 Conference. There are
no occurrences of the phrase "this reader" in the 64,109 words
in the 316 paper review forms from the 86 peer reviewers.
Reviewer X uses this phrase in section 3 of his review:
- This reader was confused by the mismatch between
the phrase "Three dimensional broom balancing problem" in the
title and the description of the problem in section 4.0.
(Emphasis added).
This phrase also appears in the signed, non-anonymous review to the MIT
Press of my first book (December 10, 1991) written by John Grefenstette
of Code 5514 of the Naval Research Laboratory in Washington:
- ... without making the intended point clear (at least to this
reader).
Here's the frequency of occurrence of the phrase "this reader"
in 7 review documents:
- Reviewer A - 0 times
- Reviewer B - 0 times
- Reviewer Y - 0 times
- Reviewer #1 - 0 times
- Reviewer #2 - 0 times
- Reviewer #3 - 0 times
- Reviewer T1 - 0 times
- Reviewer T2 - 0 times
- Reviewer T3 - 0 times
316 reviews by 86 GP-96 reviewers - 0 times
Of course, the use of a self-referential style that is not commonly used
by scientific peer reviewers does not alone establish the identity of the
writer of reviews X.
7.6. Reviews A, B, #2, and the signed non-anonymous review to the MIT
Press of my first book all used the quaint and infrequently used word "Studies"
The somewhat quaint and infrequently used word "studies" appears
in 4 of the review documents.
The words that are in more common usage are "work," "research,"
"paper," "article," etc.
This quaint and infrequently used word appeared
- 1 time in review A of my MLC paper on empirical discovery,
- 1 time in review B of my MLC paper on empirical discovery (for which
the time sequence of the plagiarism placed as being written using review
A as a template)
- 2 times in review #2 of my paper submitted to the Evolutionary Computation
journal, and
- 1 time in the signed, non-anonymous review to the MIT Press of my first
book (December 10, 1991) by John Grefenstette of Code 5514 of the Naval
Research Laboratory in Washington, DC.
How quaint and unusual is this word among contemporary peer reviews in this
field of science?
Referring to the computer file containing the 64,109 words of the 316 paper
review forms from the 86 peer reviewers of the genetic programming papers
at the Genetic Programming 1996 Conference, we found that the frequency
of occurrence of six alternative words contained in this corpus of writing
by peer reviewers in this field:
- article --- 11
- literature --- 34
- paper - 734
- research --- 46
- studies --- 1
- work --- 220
Of course, the use of a quaint and uncommon word does not alone establish
the identity of the plagiarizers involved here.
This is not the only occasion when we encounter two peer reviewers of the
same paper both belonging to the small minority of peer reviewers who use
this particular quaint and infrequently used word. See
section 3.7.
7.7. Reviews B and X and the signed non-anonymous review to the MIT
Press of my first book used the forceful and infrequently used word "Strengthen"
Most scientific peer reviewers use a rather passive and bland vocabulary
even though most English teachers advocate the use of forceful and colorful
words.
Most writers have favorite words that they use (and overuse) in their writings.
We again make reference to the computer file containing the 64,109 words
of the 316 paper review forms from the 86 peer reviewers of the genetic
programming papers at the Genetic Programming 1996 Conference.
The forceful and colorful verb "strengthen" does not appear at
all in the computer file containing the 64, 109 words.
However, this particular word does appear in each of the following 4 very
short documents:
- review B of my MLC paper on empirical discovery,
- review X of my MLC paper on optimal control strategies,
- review #2 of my paper submitted to the Evolutionary Computation
journal, and
- the signed, non-anonymous review to the MIT Press of my first book (December
10, 1991) by John Grefenstette of Code 5514 of the Naval Research Laboratory
in Washington, DC.
Of course, the use of a particular forceful colorful word that is not commonly
used by scientific peer reviewers does not alone establish the identity
of the plagiarizers involved here.
7.8. Both reviews A, X, and the signed non-anonymous review to the MIT
Press of my first book share certain antagonistic words
We previously noted that both reviewers A and B of my MLC paper on empirical
discovery contained several occurrences of rarely used hostile words. See
Section 2.13.
We also previously noted that reviewers #1, #2, and #3 of my paper submitted
to the Evolutionary Computation journal also contained several occurrences
of rarely used hostile words. See
Section 4.8.
In the above cases, the appearance of the same hostile words in the short
reviews for the same submitted paper indicated that these words may have
migrated from one review document to another during the plagiarism process.
In this section, we suggest that the common appearance of some of these
same rarely used hostile words in two reviews for different papers may suggest,
to a small degree, the possibility that the same person was the reviewer
for more than one paper.
When we searched the computer file containing the 64,109 words of the 316
paper review forms from the 86 peer reviewers of the genetic programming
papers at the Genetic Programming 1996 Conference, we found the following
frequency of occurrence for the following words:
- suspicious --- 0
- claimed (or claim or claims) --- 4
- (who cares?) --- 1
- irrelevant (or irrelevancies) --- 0
- annoyingly (or annoy or annoying) --- 0
Reviewer A of my MLC paper on empirical discovery says,
- The presentation suffers from an abundance of irrelevant
details
Reviewer X of my MLC paper on optimal control strategies says,
- The presentation suffers from the presence of irrelevancies
Reviewer A of my MLC paper on empirical discovery says,
- results are claimed to appear
The word "claimed" also appears 2 times in the signed, non-anonymous
review to the MIT Press of my first book (December 10, 1991) written by
John Grefenstette of Code 5514 of the Naval Research Laboratory in Washington:
- The claim that
John Grefenstette also mentions,
- claims of significance
Of course, the use of several different hostile words that are not commonly
used by scientific peer reviewers does not alone establish the identity
of the plagiarizers involved here.
Author: John R. Koza
E-Mail: NRLgate@cris.com
Go to top of NRlgate Home Page
Go to Abbreviated Table of Contents
Go to Detailed Table of Contents
Go to Complaint Letter
to the Evolutionary Computation journal
Go to top of Previous Page
Go to top of Next Page