Sunday, November 1, 2015

[Yasmin_discussions] why scientists engage in art science practices


francois joseph lapointe refers us to his article that argues that
there is a large (20%) cohort
of scientist-artist hybrids - and this in part addresses my point
about the danger of an
artificial art vs science divide- and is part of the argumentation
that robert root bernstein
has been developing by documenting the art avocations of the most
successful scientists
and engineers

roger malina
to see the figures you have to look up the article on line


François-Joseph Lapointe, Département de sciences biolo- giques,
Université de Montréal, Québec, Canada.

E-mail: <>.

See <>for supplemental files
associated with this issue.


Much has been said and written about the two-culture paradigm
separating the world between artists and scientists. On one side of
this debate are those who accept this cultural art/science divide. On
the other side are those who reject it altogether to promote a better
integration of artscience practices. In this paper, the author
presents a network analysis of 40 papers submitted to the SEAD Network
for Science, Engineering, Arts and Design and tests the hypothesis
that the papers submitted by artists and scientists are disconnected
in the corre- sponding graph, as predicted by the art/science
separation. Rejecting this hy- pothesis will provide support for the
alternative artscience integration.

In this meta-analysis of the 40 White Papers submitted to the SEAD
Network, I relied on network analysis, employing a wide range of
statistical and graph-theoretical approaches simi- lar to those I use
routinely for the analysis of gene similarity data [1]. I intended to
look at the two cultures from an objec- tive standpoint, testing the
corresponding hypothesis that texts authored by artists/scientists
would be separated in a network representation of intertextual
distances. In other words, if it is true that artists and scientists
think differently, the papers au- thored by artists and scientists
should fall in different clusters in the network, with papers
co-authored by both artists and scientists falling in between.

Materials and Methods

Pairwise intertextual (lexicometric) distances [2] were comput- ed
among all papers based on presence/absence data of words. The 40x40
distance matrix was then submitted to a hierarchical clustering
algorithm to identify relevant groups in the dataset using Ward's
criterion [3], and the resulting dendrogram was used as a template to
identify significant clusters. A network was also built from the
distance matrix converted into an adja- cency matrix using a 10%
resemblance threshold as a cutoff value for building the graph.

The nodes were colored in the final graph with respect to the three
different categories of papers. The texts were coded as either
"artist" or "scientist" depending of the self-proclaimed status of the
authors, or as "artscientist" for texts with authors with a dual

A statistical evaluation of network indices (degree distribu- tion,
clustering coefficient, diameter, and density) associated

with each category of papers was performed, and significance

scientists (black) should be more densely connected within each
category than among categories. Moreover, the artscientists (grey)
should be equally connected to nodes rep- resenting either one of the
other two categories.

Results of all pairwise Mann-Whitney tests were not signifi- cant, but
examination of actual values indicates higher degree, clustering
coefficient, and density values as well as a smaller diameter for
papers assigned to the artscientist category with respect to the
artist and scientist categories. When these test statistics were
compared to those obtained from 1000 random graphs, the artscientist
was the only category with observed values more extreme than those
expected by chance alone. In other words, whereas all network indices
could not discrimi- nate between texts authored by artists or by
scientists, the pa- pers authored by artscientists were clearly more
clustered and connected to each other, as well as the other two

The White Papers have spoken: contrary to Snow's para- digm [5]
artists and scientists are not distinguishable, but artscientists are
a different breed – individuals who thrive in transdisciplinary
contexts. This is, of course, what the SEAD Network is all about, but
this meta-analysis provides statistical support for promoting such
collaborative endeavors.

Fig. 1. Joint representation of the clustering and network analysis of
the 40 White Papers submitted to the SEAD Network based on
intertextual distances. The different colors associated with the four
clusters defined on the dendrogram are used to identify the
corresponding nodes in the network. The three categories of pa- pers
are also identified on the graph by nodes with white (artists), black
(scientists) or hatched (artscientists) labels. (©François- Joseph

References and Notes

*This paper was presented as a contributed talk at Arts, Humanities, and Com-


of the corresponding statistics was assessed with pairwise

plex Networks – 4

Leonardo satellite symposium at NetSci2013. See

Mann-Whitney tests among the various categories. These val- ues were
also evaluated with respect to a random graph model [4] in which all
nodes have the same probability of being con- nected.

Results and Discussion

The hierarchical classification of intertextual distances is pre-
sented alongside the network analysis of the 40 White Papers in Figure
1. That joint representation identifies four clusters in the
dendrogram corresponding to partly overlapping subgraphs in the
associated network. More interestingly for the purpose of this
meta-analysis is the clustering of nodes representing the three
different categories of papers. Under the art/science sepa- ration
hypothesis, nodes associated with artists (white) and


1. J. Beauregard-Racine, C. Bicep, K. Schliep, P. Lopez, F.-J.
Lapointe and E. Bapteste, "Of woods and webs: Possible alternatives to
the tree of life for studying genomic fluidity in E. coli," Biology
Direct 6 (2011) p. 39.

2. C. Labbé and D. Labbé, "Inter-textual distance and authorship
attribution Corneille and Molière," Journal of Quantitative
Linguistics 8 (2001) pp. 213- 231.

3. J. H. Ward, Jr, "Hierarchical grouping to optimize an objective function,"

Journal of the American Statistical Association 58 (1963) pp. 236-244.

4. P. ErdÅ‘s and A Rényi, "On random graphs," Publicationes Mathematicae 6

(1959) pp. 290-297.

5. C. P. Snow. The Two Cultures: And a Second Look. (Cambridge, U.K.:
Cam- bridge Univ. Press, 1964).

©2014 ISAST doi:10.1162/LEON_a_00780

LEONARDO, Vol. 47, No. 3, p. 277, 2014


Yasmin_discussions mailing list

Yasmin URL:

SBSCRIBE: click on the link to the list you wish to subscribe to. In the page that will appear ("info page"), enter e-mail address, name, and password in the fields found further down the page.
HOW TO UNSUBSCRIBE: on the info page, scroll all the way down and enter your e-mail address in the last field. Enter password if asked. Click on the unsubscribe button on the page that will appear ("options page").
TO ENABLE / DISABLE DIGEST MODE: in the options page, find the "Set Digest Mode" option and set it to either on or off.
If you prefer to read the posts on a blog go to