Eldling
Eldling
Eldling
MATHEMATICAL SOCIOLOGISTS ON
MATHEMATICAL SOCIOLOGY
Christofer Edling
1'ntroduction
1
1 wish to thank the .cholars who accepted not only to be interviewed by me, but
who al o shared valuable experiences and suggestions; Peter Abell, Philip Bonacich
Kathleen Carley, Patrick Doreian, Thoma Fararo and Harrison White. 1 owe thc
late Aage Sorensen for his enthusiasm in this project and Ior sharing many
thoughts in extended discussions. Part of this work has been presentcd at Utrecht
University and at Swedish CoUegium for Advanced Study (SCAS) and 1 thank
participants in those seminars for useful remarks and suggestions. In addition, l wish
to thank Peter Hcdstrorn Frcdrik Liljcros, Wcrner Raub and Richard S, cdberg
for vcry helpful suggestion ·.
346 WE ALWAYS KNOW CHRISTOFER
MORE THAN WE CAN 346
there with no technical background I'm not sure the gulf could have
been crossed" (Waldrop 1992: 195). Thirdly, because it is a unifying
language, mathematics provides a way to learn from other scientific
domains that struggle with structurally similar properties. Mathematics
can bring our attention both to the isomorphism between sociology and
other sciences and by way of example suggest possible ways to apply
mathematics in sociology (Coleman 1964; White 1997).
From social network analysis we learn that mathematics can serve the
social sciences in just the same way as it serves the natural sciences. In
the development of social network analysis mathematics has provided
a tool to tackle complexity, a language to ease communication, and a
way to bridge theoretical concepts in sociology, mathematics, anthropol
ogy, computer science, physics, etc. So we do know that mathematics
can help us make considerable scientific progress. However the rest
of the chapter takes as a point of departure the fact that by and large
mathematics is rarely put to work in sociology and when it is, it is done
by a small number of sociologists.
lntertnetus
While both Patrik Doreian and Harrison White did identify stronger
with the label mathematical sociologist they too did throw in caveats.
Doreian, long term and highly dedicated editor ofJoumal efMathematical
Sociology and since recently the editor of Social Netuiorks said that,
349 WE ALWAYS KNOW CHRISTOFER
MORE THAN WE CAN 349
Givcn that I work in a lot of different substantive areas, and that I always
use some mathematics at sorne point I would havc to say it i
math ematical sociology. When I describe what I do, often I use
mathematical sociology as a description. I have tried using "playing in a
mathcmatical play pen ', but that generates a similar response to saying
mathematical ociology. Maybe the conjunction of play' and
'mathematical' was the problern! (Patrik Doreian)
that this tension in itself is really the core of sociology, as sorne have
clairned (Lepenies 1988). Whatever is the case; with sorne confidence
I would say that a mathernatical sociologist certainly subscribes to
the idea that sociology is a science of the social possibly to the idea
that it can be both. Of course this <loes not help rnuch as sociology
has not decided what it rneans to be a science of the social. But nor,
do I believe have rnathernatical sociologists. Just as sociology at
large mathernatical sociology is a very heterogeneous field (Freese
1980). What <loes constitute its borders is a shared conviction that
rnathernatics can bring leverage to sociological analysis. Even though
this might be debated, most sociologists and including the ones I have
interviewed,
would make a very clear distinction between statistical modeling
and mathernatical sociology. Indeed the literature on mathernatical
sociology and the interviews with mathernatical sociologists suggest
that rnathematical sociology is strictly a theoretical affair. Thus, it is
the application of mathematics to sociological theorizing that defines
mathernatical sociology. And mathernatics is used broadly to include
also logic and cornputer models. Formal theory could therefore be an
alternative label and is often used as such. However, it would be at fault
not to acknowledge that statistical modeling is the area within sociology
in which mathernatics has the strongest irnpact on the field as a whole.
The statistical tools being put to work in network analysis, eventhistory
analysis, and hierarchical rnodeling do bring added sophistication to
quantitative sociology.
Nevertheless, classifications of rnathernatical sociology highlight the
use of mathematics for constructing theoretical rnodels of social phe
nomena. Even though many rnathernatical sociologists are using and
ometirnes even develop quantitative methods they often point out
explicitly that the use of rnathernatics in sociology should not be equated
with statistics. And the people I interviewed all subscribe to the idea
that rnathernatical sociology belongs to theoretical sociology proper. In
g neral there is no agreernent on what is theoretical sociology. Many
things indeed go into that label. In this respect mathernatical sociolo
gists seem to be a bit more united than the discipline at large.
Many including myself take the view that theoretical model building
is an act of balancing realism, generality, and precision. In this act of
balance, one will have to stand back in favor of the other two sirnply
because ali things cannot be achieved (Heckathorn 1984; Levins 1966).
If one accepts this prernise, one can continue discussing the pros and
cons of rnodel building. Mathematical sociology is often accused of
351 WE ALWAYS KNOW CHRI
MORETOFER
THAN WE CAN 351
of the unrealistic assumption will cancel each other that small devia
tions from realism result in mall deviations in the conclusions, and
that in any case, the way in which nature departs from theory will
uggest where future complications will be useful" (Levins 1966: 422).
To illustrate the balance of realism generality and precision I wish to
give two examples of simple mathematical rnodels that successfully walk
the line. Incidentally both model are submitted by non ociologi ts
although Watts later transferred to a sociology department. Schelling's
model of segregation i a classic example and Watts' and Strogatz'
model of the small world is on its way to becoming one.
The motivation for Thomas Schelling's dynamic models of segrega
tion was the observation that people get segregated across many dimen
ions and often by discrimatory individual behavior. "By 'discrimatory'
I mean reflecting an awarcness, conscious, or unconscious, of sex or age
or religion or color or whatever the basis of segregation is, an aware
ness that influences decisions on where to live, whom to sit by what
occupation to join or to avoid whom to play with or whom to talk to"
(Schelling 1971: 144). But ethnic segregation was the prime concern
at the time when Schelling proposed his model, and it is primarily in
uch light that it has been di cus ed. Yet, Schelling is very clear that
the model as such is abstract enough that results can be generalized to
any preference driven segregation process. Thus following Fararo the
model is fueled by concern for a real social issue and in that sense i t is
a realistic model. But it is also a general model claiming to shed
light on a range of segregation phenomena. However Schelling
remark early on that sorne very important factors that explain
egregation are not included such as in titutionalized segregation
(that he calls orga nizational) and economically driven segregation. In
Schelling s models the egregation process i driven by individual
preferences only. In this re pect they are also fairly precise models.
Both initial and boundary conditions are clearly defined but because
the analysis was made by hand agents' movement rules for instance
are only loosely defined (but contemporary computer replications give
exactly the sanie results).
One of Schelling's models is a two dimensional lattice of 13 rows
and 16 columns on which 25 starchips and 18 zerochips are distrib
uted randomly. Assuming that both stars and zeros have a preference
for having at least 50 percent neighbors of their own type Schelling
rearranges the chips so that this condition is met for every chip. The
result is a strikingly segregated pattern in which a star have about
80
353 CIIRISTOFER
WE ALWAYS KNOW MORE THAN WE CAN 353
o # # # # o o o o # # o o ### 000 # # oo
o # # # # o o o o # # o # # O O # # # 00
# # o o # # # # # o # # # # # ####### #00
# # # o # # # o o ## ### ###00
o o o # # # # # # o o #O O #####
000
# o # o o # o o # # 0000# #0000## # #
# o o # o o o # # # 00000 # o o o # #
o # o # # # o o o # o o o # o o o o #
o # o # # o # o o## ##o o o #
o o # 0#0000## 00#### ##0000 ##
o # # o o o o o # # o # # 0###000 # # o # #
# o # o # o o # o # o o o o o 00###00 #
o o o # o # o o o # # 00 o 000 o #
Fig. 7 Fig. 9
percent of her neighbors being stars and a zero has about 83 percent
zeros in her neighborhood (in the random initial clistribution the cor
responcling numbers was 46 and 53 percent).
Schelling's dynamic model of segregation is in many ways a role
model in mathematical sociology. Not only does it strike the balance
between realism, generality, and precision perfectly but it also generates
surprising, and even counterintuitive results. Recall that in the model
segregation emerges solely from the action of uncoordinated individual
actor and in that respect this model is a beautiful example of the link
between microinteraction and macrodynamics.
The rationale far Duncan Watts' and Steven Strogatz's model of
small world dynamics (Watts and Strogatz 1998) is not motivated by
a pressing social issue such as segregation. Rather it is the answer to a
puzzling question of how to understand a particular network structurc.
Far a long time the ideal type networks that mathematical
sociologists worked with were ei ther random networks or ordered
networks despite the fact that all sociologists knew empirically that
social networks are neither randorn nor ordered. For one, social
networks are clustered that is, people tend to form tight knit groups
that are clistinguishable from other groups. Trying to capture this
intuition mathematically for instance by modeling biased networks is
hard. A particularly interesting and popularized network
communication phenornenon is the socalled smallworld effect. This
effect was illustrated experimentally in the
l 960s with the implication that only six intermecliary steps separate any
two Americans. A question that had deluded sociologists since these
354 WE ALWAYS KNOW CHRISTOFER
MORE THAN WE CAN 354
experiments and that captured Watts' imagination was this: what makes
a network small? In Watts' model actors are put in an ordered circle
and every actor is connected to its four nearest neighbors. One of the
actors is elected and at random one of its connections reconnected
to another randomly elected actor in the ring. As the process of re
connection i repeated the ring transforms from an ordered network
structure into a random network structure as in Figure 2 (Watts and
Strogatz 1998).
Starting from a very simple model of an ordered network structure
Watts could demonstrate that what is characteristic of a smallworld
network is both a high level of clustering and a short average distance
between actors in the network. Thus a smallworld network seems to
capture both the property of random networks, i.e. short path lengths
and ordered network i.e. high clustering. What is really interesting
with Watt's model is that parameter p, the degree of randomness
need only be very small for the ordered structure to transforrn into a
smallworld network.
Watts smallworld model is even simpler than Schelling's model
of segregation but still very elegant. In terms of the balance between
realism generality and precision, the judgment is perhaps not as clear
cu t. The model as such is not particularly realistic in the sense that it
<loes not even abstract from observed social reality. Few sociologists
would feel at home at this level of abstraction. On the other hand,
p= o p=l
increasing randomness
the model is general and has applications way beyond sociology; the
network properties under analysis are present in many biological and
physical networks, both natural and designed. And it is precise and
welldefined. In contrast to Schelling's model this one needs a fair deal
of interpretation before it can be mapped onto social reality. But it
does highlight important properties in alrnost ali social networks that
are now, because of this model, easier to understand and theoretically
scrutinize.
These two examples do bring out both the attractions and the poten
cial shortcomings of mathematical sociology that I believe are nicely
captured by the balance of realism, generality, and precision. Another
way to phrase this problem is in terms of a conflict between sociological
intuition and formal theorizing, as Fararo does in this quotation.
I alway think about it, as we always know more than we can ay. And
we always can say more than we can really forrnally put clown in more
exacting terrns. So as you go further and further frorn the fundamental
intuitions in the interest of being logical and rnathernatical, you can
potentially loose contact with the governing intuitions. But the main
gain would be to try to bring the rnathernatics back into, and as close
as possible to the basic intuitions of the field. Trying to represent thosc
intuitions in sorne way. lt is what Bourdieu calls a habitus. It is a kind of
a sociological habitus you acquire by exposing yourself to classical ideas
and postclassical idea and so on. You know, you think sociologically and
then you think rnathernatically. But these are often hard to fit together.
The mathernatics enforces a discipline that the other discipline does not
really value in the same way. It has it its own forms of rigor but they are
not the sarne. To bring those two into conjunction has always been the
sort of thing that I thought of as irnportant. I do not think we've been
that successful really, but collectively we are trying (Thornas Fararo).
Doreian also draws our attention to the third type of models, models
of action. Prototypical examples of such models would be those of util
ity maximizing individuals, lndeed James Coleman proposed that the
simple and well defined assumptions about individual racional choice
should provide the foundation for social theory (Coleman 1990). This is
not yet the case, neither in sociology at large (but see Goldthorpe 2007)
nor in mathematical sociology. On the contrary, if there is one thing
that n1y earlier review (Edling 2002) clearly shows it is that mathernati
cal sociology spans across a broad range of phenomena and applies an
equally broad set of mathernatical tools to try and understand those
phenomena.
358 WE ALWAYS KNOW CHRISTOFER
MORE THAN WE CAN 358
because this is how their training wired them to think. But people that
start out in social science and decide to bring in mathematics and other
formal tools actually turn away from the highway and pick the straight
and narrow road. I think Abell is right in his analysis that the racional
choice in contemporary sociology is to not do formal theory.
I think at sorne stage ·ociology just has to fa.ce up to that. Sociology is
so cliverse and because it shades of into history, there will always be this
other end of things. But it has to change. 1 am not optimistic about it
changing in the foreseeable future, because I think the present "social
construction" of sociology is such that there is a mutual interest between
teachers tudents, publishers, cte. to maintain the present paracligm. In
my vicw it is intellcctually ub optimal. Neverthclcs it is cquilibrium.
Everybody wants it and everybody in it would be worsc off by moving
away from it. So you find it vcry clifficult to break that. Human history
is littercd with failed intellectual tradition. Not empirical sociology, but
social theory is a failed intellectual tradition, and future generations will
necessarily see it as such. At the moment, though, it is relatively stable
(Peter Abell).
but Doreian sirnply rests his case, <loes his thing, and leaves it for oth
ers to decide on the utility of what he is doing. Peter Abell, who also
carne from science to sociology hares with the social science group
a strong belief in the value of rnathematics. He actually argues that
ocia! scientists are rnorally obliged to be "clear and precise" in their
thinking, and that mathematics is needed to guarantee this.
Sorne of the interviewees also cornrnented specifically on weaknesses
and potencial risks of mathernatical sociology. Peter Abell for instance
compares sociology to economics and warns that in economics the rnath
ernatics is sornetimes valued so highly that it overshadows the substancial
issues. He argues in effect that our curiosity of empirical cornplexity
always should have the upper hand over technical proficiency in defin
ing what problerns to study. In the same line of reasoning, Harrison
White points out that even if the rnathematical approach by itself is
praiseworthy we rnust be aware that the mathematics we need might
not actually be the mathematics we have access to at the moment. As
a result, sociologists must have a flexible attitude towards mathemat
ics and, which is of course rnuch, rnuch harder, sociologists rnust keep
a constant eye on developrnent in mathematics. Looking at the sarne
issue frorn another angle, Thornas Fararo sees a problem in bringing
rnathematics and sociology together. To his rnind they represent two
diflerent rnodes of thinking, two intuitions, which are not always easy
to link. The risk arises when one continues to construct rnathernatical
rnodels even when the mathernatical and the sociological intuitions are
not in touch with each other. This leads to the sarne problern that Abell
identified in economics the problern which arises when the models no
longer represent the sociological intuitions.
Concluding Remarks
this point. My interviews bear witness that there are positive adherents
of mathematical sociology, and my literature review (Edling 2002)
provides references to impressing work on many timely problems. At
the same time, it is obvious that mathematics has not established itself
as a general tool in the discipline.
Several people interested in mathematical sociology testify to the
institutional fragility of mathematical sociology, evident in many ways,
and most likely derived from the small number of adherents of this sub
discipline. For example, both Peter Abell and Patrick Doreian got into
mathematical sociology through a program at U niversity of Essex. This
led me to believe that there was actually something exciting going on
in mathematical sociology in England in the late l 960s. Apparently, this
was not the case. Apart from these two, no lasting contribution to the
field evolved out of Essex. Doreian's unconcerned and laconic observa
tion that the program "died" leads me to conclude that although there
where certainly an interpersonal xenogamy of ideas going on (Abell
1971; Doreian 1970) the program in itself is not much to be spoken of.
Indeed, Doreian left for a position in US in the early l 970s and Abell,
who remained in England, become devoutly disillusioned of the future
of any version of mathematical application in sociology.
Patrick Doriean was the editor of the Joumal of Mathematical Sociology
between 1982 and 2005. lt is beyond doubt that the journal has sur
vived until this day much through his hard work. While this has saved
the journal from going under, it is possible that, by attaching survival
chances to the energy and professional contact network of one single
individual, its vulnerability has also increased. Again, this is the problem
of a small field. A parallel case is the Internacional Network for Social
Network Analysts (INSNA). With the growing popularity of network
analysis, INSNA has become a firmly established organization and is
no longer dependent on one dedicated person, but for a long time it
almost was. But, according to Doreian's account, even after over 30
years since its inception the Journal of Mathematical Sociology struggles
to
stay alive. A small field simply cannot support a specialist journal, and
especially so at times when institucional subscription rates are skyhigh
and academic libraries experience budget shrinkages year after year.
It is not the case that there are no outlets for mathematical sociology,
severa! papers in general sociology journals draw on mathernatical
or computational modeling. Indeed, this is a very positive thing for
mathematical sociology.
364 WE AL WAYS KNOW CHRISTOFER
MORE THAN WE CAN 364
Rejerences
Dorcian P. 1970. Mathematics and tite 'tudv q/ acial Relations London: Weideníeld and
Nicol on.
Edling, C. R. 2002. Mathcrnati .s in ociology" in Annual Review oJ ociology,
28: 197 220.
Fararo, T. J. 1989. 77Le Meaning ef General Theoretical . 'ociology: Tradition and Formalization,
Cambridge: Cambridge Univcr ity Pre s.
. 1997. "Reflections on mathernatical sociology" in Sociological Forum, 12: 731 O 1.
Freeman, L. C. 1984. "Turning a profit frorn mathematic : The case oí social nerworks"
in Joumal ef Mathematical 'ociology, 1 O: 343360 .
. 2004. The Deoelopment of . acial Netuiork Analysis: A Study in tite . ociology of aence.
Vancouver: Empirical Pre .
Free e, L. 1980. 'Formal theorizing' in Annual Review of 'ociology, 6: 187 212.
Goldthorpe J. H. 2007. On ociology vol. 1, Second edition Stanford: Stanford Uni
ver ity Press.
Hannan, 1\11. T. andJ. Freeman. 1989. Organiratumal Ecologv Cambridge MA: Harvard
University Pres .
Harary I· ., R. Z. Norman and O. Cartwright. 1965. Structural Models: An Introductioti
to tite Theory of Dlrected Graphs Wilcy: Ncw York.
Hcckathorn, D. D. 1984. "Mathcmatical thcory construction in sociology: Analytical
power, scope, and descriptive accuracy as tradeoff" in JoumaL ofMathematical Sociol
ogy, 1 O: 295323.
Hcdstrom, P. 2005. Dissecting the Social: On tite Principies ef analytical sociology, Cambridge:
Cambridge University Press.
Helbing O. 1994. "A mathematical model for the behavior oí individuaJs in a social
ficld' in Journal ef Mathematical Sociology, 19: 189219.
Lcp nie , W. 1988. Bettueen literature and science: The rise of sociology Cambridge: Cam
bridge University Press.
Lcvins, R. 1966. The strategy of model building in population biology" in American
. 'cientist 54: 421431.
Macy, M. VV. and R. Willer. 2002. From [actors to actor : ComputaLional sociology
and agentbased modeLing' in Aunual Review ef Sociology, 28: 143 166.
Maynard Smith J. 1982. Eoolution and tite theor» cfgames Cambridge: Cambridge Uní
vcrsity Prcss.
ash, J. 1951. 'Noncooperative gamc in Annals ef Mathematics 54: 286 295.
cwman �1. E. J., A. L. Barabási and D. J. Watts. 2006. Tii.e structure and dynamics ef
netuiorks, Princeton, [J: Princcton Uni ersity Pre s.
chclling, T. C. 1971. "Dynamic model oí segregation in Joumal of Mathematical
Sociology, 1: 143186.
oren en, A. B. 1978. 'Mathernaticalrnodel in ociology in Annual Rana» ef ociology
4: 345371.
upp , P. 195 7. Introduction lo logic, Princeton, NJ: D. Van No trand Co.
wedberg, R. J 990. Economics and sociology, Princeton NJ: Princeton Univer ity Pre s .
. 2001. Sociology and gamc thcory: Contemporary and hi torical per
pectives in 77zeory and Sociery 30: 301 335.
\ aldrop, M. M. 1992. Complexuy: Tlie etnerging science at tlte edge ef order and chaos, New
York: Simon & Schustcr.
Watts D. J. 2003. Six degrees: Tiie science ef a connected age, New York: Norton .
. 2004. "Thc 'ncw' sciencc of nctworks" in Annual Review of Sociology; 30: 24·3270.
'Watts, D.J. and S. H. Strogatz. 1998. "Collective dynamics oí'smallworld' networks'
in. 'ature, 393: 440442.
Wcidlich, W. 1994. "Synergetic modeling conccpts for sociodynamics with applica
tion to collective political opinion formation" in Joumal ef Mathematical ociology,
18: 267291.
367 WE AL WAYS KNOW CHRISTOFER
MORE THAN WE CAN 367
Wellman B. 1988. 'StructuraJ anaJysis: From method and metaphor to theory and
substance" in B. Wellman and S. D. Berkowitz (eds.), Social structures, Cambridge:
Cambridge University Press, 1961.
White, H. C. 1997. "Can mathematics be. ociaJ? Flexible representations for interaction
process and its sociocuJturaJ constructions ' in Sociological Forum, 12: 537 l.
Von Neumann, J. and O. Morgenstern. 1944. Theory oJ gam.es and economic beluunor
Princeton: Princeton University Press.