I followed with interest the recent discussion about Gong et al's inclusion
of conlangs in their paper, and the question it raised about their knowledge
(or lack therof) of the conlangs they cite, the conlang community's
responsibility to provide warnings about the fictional nature of the hobby,
and the use of conlangs as evidence in a research paper.

I then read through Gong et al's paper with this question in mind:  does the
fact that conlangs were included as examples constitute a serious mistake, a
harmless error, or a legitimate use of evidence?  I decided that any
linguistic research that examines the development of natlang structure
(syntax, phonology, grammar, etc) should not use a conlang as an example. 
This is because, all else being the same, conlangs evolve through conscious
creation, whereas natlangs evolve through unconscious collaboration between
a community of speakers.  Therefore, the processes by which conlangs and
natlangs gain their structure are fundamentally different.   However, any
research that seeks to understand how a language might operate, or how a
particular system of grammar can work in practice, should be able to also
examine conlangs since many (I'm willing to wager the majority) of conlangs
are, in fact, usable.

So how does Gong et al's paper stack up?  My assertion is that they made a
serious mistake. Gong et al constructed a model to test the evolution of
word order by simulating rule acquisition between speakers.  This is an
example of natural language evolution, and as such is very different from
the development and implementation of conlangs as I said above.  

I don't think this  invalidates the conclusions of the paper as a whole,
because the section that mistakenly mentions conlangs is not part of the
main focus of the research.  Furthermore, they state that the conlangs as
members of a very small community of ergative VSO languages; removing the
conlangs, or recognizing that the conlangs are invalid evidence, actually
strengthens the paper's assertion that ergative VSO languages are very rare.

I noticed that earlier in the paragraph mentioning conlangs, the authors
cite Dryer and Comrie.  It would be interesting to see if these authors also
included conlangs in their work, or not, but either way, Gong et al's
statement mentioning conlangs does not cite any sources; one wonders how the
authors found their information.  A review process should have picked out
the lack of citation at the very least, I would think.

My conclusion:  The conlang/natlang mistake and lack of citation makes me
wonder how many other things in this paper were poorly investigated or
improperly cited.  I do believe the paper is valuable, but I take its
conclusions with a grain of salt and would be very reluctant to cite it in
my own publication.

Due to the complexity of the model and the generally professional tone of
the paper, I seriously doubt the intentional inclusion of conlangs as
evidence would be their idea of a joke. Intentionally adding misleading data
is a serious charge in science. vI would never allow myself or my (future)
students to engage in this activity in professional research.  

I share the opinion expressed on the previous thread that conlangers do not
have to go to extremes to demonstrate the fictional nature of their
creations.  Citations in academic research should be based on more rigorous
sources than Wikipedia/Frathwiki, or personal web pages.  If I make a
website on Angosey and happen not to mention that it's a conlang, but then
turn around and find it cited in academic research without even being
contacted by the authors, I will not feel like I was party to a deliberate
deception. Rather, the authors never bothered to check preexisting
literature to determine if Angosey was in fact real.  Furthermore, even if a
conlanger intentionally attempts to deceive the linguistic community, a
quick source check would reveal that the conlanger was not a reputable
source for natural language descriptions (unless a member of the linguistic
community attempts to hoodwink other members, which I admit could work).

If there is a place for conlangs in linguistic research, I imagine it would
be limited.  Two possibilities come to mind: either the exploration of the
use and implications of certain structures, or the ability of an individual
to approximate the communal process of building a language.  There may be
more that I am not thinking of.

However, we seldom study sculpture in order to understand the human body, or
examine music to determine the properties of sound.  I expect the science of
linguistics to follow the same pattern regarding the connection between
constructed and natural languages. 

Danny