PERSPECTIVE
The golden age of social science
Anastasia Buyalskaya
a,1
, Marcos Gallo
a
, and Colin F. Camerer
a,b
Edited by Matthew O. Jackson, Stanford University, Stanford, CA, and approved November 23, 2020 (received for review May 14, 2020)
Social science is entering a golden age, marked by the confluence of explosive growth in new data and
analytic methods, interdisciplinary approaches, and a recognition that these ingredients are necessary to
solve the more challenging problems facing our world. We discuss how developing a
“
lingua franca
”
can
encourage more interdisciplinary research, providing two case studies (social networks and behavioral
economics) to illustrate this theme. Several exemplar studies from the past 12 y are also provided. We
conclude by addressing the challenges that accompany these positive trends, such as career incentives and
the search for unifying frameworks, and associated best practices that can be employed in response.
interdisciplinarity
|
diverse teams
|
new data
|
difficult challenges
Social science is entering a golden age (1). A rise in
interdisciplinary teams working together to address
pressing social challenges, leveraging the explosive
growth of available data and computational power,
defines this moment. Each of these trends has been
written about individually
—
the
“
big data revolution
”
has been transforming social science for several years
(1), and the benefits of diverse teams are increasingly
recognized and quantified (2, 3). We argue that it is
the confluence of data, diverse teams, and difficult
challenges which makes this a unique and exciting
time for social scientists to tackle important research
questions. Of course, there have been large team ef-
forts in previous decades (4), but their frequency and
breadth have increased recently.
Funding agencies have, in turn, recognized the
need to support interdiscipl
inary teams. Fig. 1 presents
evidence from multiinvestig
ator grants funded by the
NSF of how interdisciplinary research is on the rise in
social science. Given the difficulty in defining interdisci-
plinary work, federal agenc
ieshavechosentousethe
number of grants provided to projects with multiple
principal investigators as a proxy (5, 6). These data res-
onate with our idea of what interdisciplinarity means in
this golden age: active coll
aboration among scientists
with different training
—
meaning a diversity of perspec-
tives is influencing the research
—
as opposed to one
researcher passively borrowing ideas from other fields.
We hope our perspective will encourage scientists
to take advantage of new datasets and form diverse
collaborations to answer pressing questions. We di-
rect these ideas especially to funding agencies and
academic institutions, to convince them to provide
more funding for this type of work. Ultimately, we wish
to see an acceleration in work that addresses difficult
challenges. For instance, the COVID-19 pandemic il-
lustrates how large-scale problems will only be solved
by many scientists contributing what they know best.
The Need for a Lingua Franca
The opening of disciplinary borders is akin to an in-
creasing trade of methods, language, and knowledge
across fields. This concept of trade is built on the
premise that, like people and countries, each social
science discipline has a different endowment (i.e., a
historical mastery of tools and accumulated knowl-
edge) and comparative advantage. Defining how the
social science disciplines differ is difficult, but even a
thumbnail sketch can clarify our ideas about compar-
ative advantages and the value of trade. Hoping that
the reader will appreciate that we overemphasize
differences in fields (and ignore variation within them),
we define them as follows. Anthropology seeks to
understand cultural differences in human societies
using ethnography, unearthing physical details of
human development and exploring mathematical
a
Division of Humanities and Social Science, California Institute of Technology, Pasadena, CA 91125; and
b
Computational and Neural Systems,
California Institute of Technology, Pasadena, CA 91125
Author contributions: A.B. and C.F.C. designed research; A.B., M.G., and C.F.C. performed research; M.G. analyzed data; and A.B., M.G., and
C.F.C. wrote the paper.
The authors declare no competing interest.
This article is a PNAS Direct Submission.
This open access article is distributed under
Creative Commons Attribution-NonCommercial-NoDerivatives License 4.0 (CC BY-NC-ND)
.
1
To whom correspondence may be addressed. Email: abuyalsk@caltech.edu.
Published January 22, 2021.
PNAS
2021 Vol. 118 No. 5 e2002923118
https://doi.org/10.1073/pnas.2002923118
|
1of11
PERSPECTIVE
Downloaded at California Institute of Technology on January 22, 2021
models of coevolution of culture and genes. Economics uses
math-heavy methods to understand systemic (general equilib-
rium) outcomes of optimization of allocation of scarce resources,
particularly money, in trading goods and services. Its main
methods include theories rooted in preferences, beliefs, and
constraints and analyses of field data. Political science studies
formal systems of government, voting, juries, and law, which in-
fluence how people make consequential decisions collectively in
different systems. Ideology is a central construct, with polls and
surveys being a cornerstone method, although media and finan-
cial contributions data are increasingly used. Psychology seeks
regularity in how people think and behave, with an emphasis on
mechanisms and constructs such as memory, attention, and
emotion. The main methods are laboratory experiments and
psychometric or psychophysiological measures (though cognitive
neuroscience uses a greater variety of newer methods). Finally,
sociology investigates how the soc
ial world is created by and influ-
ences how people act in social groups at different levels of formal
and informal aggregation. General ideas about functions of social
structure are central but are not ma
thematized as in economics (e.g.,
economists might focus on allocati
ve efficiency defined mathemat-
ically while sociologists might foc
us on social reproduction of elite
success measured statisti
cally or qualitatively).
Readers may view these highly reduced descriptions of their
own fields as overly simplified, while perhaps believing that the
descriptions of the other fields are not too bad. That perception
itself illustrates why communication is a challenge for interdisci-
plinary work. Complicating trade is the fact that many words like
“
rationality,
”“
trust,
”“
discrimination,
”“
hierarchy,
”“
salience,
”
and
“
power
”
are used across the social sciences, but in different
ways. Their local meanings are understood by
“
native speakers
”
but often baffling to
“
traders
”
arriving from foreign scientific
lands. Interdisciplinarity needs a common trade language across
disciplines, a
“
lingua franca.
”
In a useful lingua franca, all disci-
plines adopt the
“
best
”
language from whichever discipline has
described an idea most effectively
. In order for teams of researchers
to effectively tackle the complex research questions of our time, they
will need to work together to build a common vocabulary that en-
hances the efficiency of the
ir trade and collaboration.
Examples of lingua franca which originated in individual dis-
ciplines include an understanding of culture from anthropology,
rational choice theory from economics, ideology from political
science, laboratory experimental methods from psychology, and
social networks from sociology. Besides these central constructs,
powerful tools for quasi-experimental causal inference
—
which
originated in psychology (8), created a boom through more so-
phisticated use of instrumental variables in economics starting in the
1990s (9), a little later in political science (10), and somewhat in
parallel in computer science an
d statistics around 1995 (11)
—
have
evolved as a methodological lingua f
ranca across the social sciences.
A useful lingua franca, one which is to be a truly unifying
framework, will need to cut through the technical jargon specific
to any one field of origin in order to be widely accepted and used.
Taking the time to build such a lingua franca will enable diverse
teams to tackle multidimensional problems and create innova-
tions for better health, wealth, and well-being (12). Drug addic-
tion, obesity, sustainability and climate change, technology-
driven changes in sociopolitical discourse,
“
fake news,
”
and
how artificial intelligence will change our world will never be fully
understood by any one discipline working alone. Instead, making
progress on these challenges will require understanding the in-
stitutional incentives, cultural norms, cognitive mechanisms, and
social network effects that create and sustain these phenomena.
Fig. 1. Single (SPI) vs. multiple (MPI) investigator awards at the NSF, 1987 to 2018. Notice the trend toward awards with more than one PI, which
the NSF considers to be the best current proxy for interdisciplinarity (6) (data source: refs. 5 and 7).
2of11
|
PNAS
Buyalskaya et al.
https://doi.org/10.1073/pnas.2002923118
The golden age of social science
Downloaded at California Institute of Technology on January 22, 2021
Interdisciplinary work has already helped make progress in fields
including poverty, health epidemics, and mental health.
Learning from Case Studies
In the next section, we present two
“
case studies
”
of successful
interdisciplinarity: social network science and behavioral eco-
nomics. In both cases, interdisciplinary research led to the crea-
tion of new cross-disciplinary fields of inquiry built on the
comparative advantages of contributing fields, inspiring a shared
lingua franca, generating insights about human nature, and im-
proving social outcomes. These cases originated decades ago, so
they are not meant to illustrate the three features that we take to
characterize the golden age. While the original research was not
particularly propelled forward by large, diverse datasets or by a
desire to tackle global challenges, recent research has moved in
those directions (Figs. 2
C
and
D
and 3
C
).
Social Networks
Social networks are our first case study of a successful interdisci-
plinary enterprise. Network analysis uses methods from physics,
computer science, and applied math to analyze questions often
studied by sociologists, anthropologists, and psychologists re-
garding how interpersonal relationships are formed and how
behaviors, beliefs, and emotions are transmitted across con-
nected individuals (13). One striking feature of network analysis is
the diversity of scholars who have been active in researching this
field from the beginning, and who continue to contribute to in-
tellectual progress (see Fig. 2 for some examples). People from
different fields, traditions, and countries have worked together on
related research questions (14). Network analysis has been sig-
nificantly enabled by the availability of novel datasets, such as
social media connections, and data from increasingly
“
con-
nected
”
devices such as fitness trackers with social aspects (15).
Notable contributors to the field of network analysis are Watts
and Strogatz (16), who brought to light several key network
properties, including that real-world networks are neither totally
ordered (there are not always clear rankings between nodes) nor
completely random (with all nodes having unequal probabilities of
being connected with other nodes). Their work was important in
getting the statistical physics community to recognize that their
techniques could be applied to social settings, thus catalyzing an
interdisciplinary turning point. It is worth noting that subsequent
research, which flourished primarily in sociology, economics, and
applied mathematics, did not necessarily follow directly from this
original paper.
One attractive feature of network science is that simple
mathematical models capture the core features of complex net-
works, allowing the study of network dynamics across a variety of
phenomena. The seemingly unrelated affiliations between actors,
power grid transmission lines, and the neural network of
Caeno-
rhabditis elegans
can all be captured via a simple
“
small-world
”
network model, a mathematical graph in which the nodes (indi-
viduals) are not neighbors with most of the other nodes and yet all
other nodes can be reached in a small number of steps (17
–
19).
Example 1: Revisiting Influence and Information Transmission.
Collective behaviors are often studied at a static point in time,
implicitly assuming that all individuals simultaneously make in-
dependent decisions. However, the heterogeneous process of
information accumulation and integration prior to decision-
making suggests that many decisions are actually made sequen-
tially and that beliefs can be
“
transmitted
”
from one individual to
the next. Given how many behaviors
—
from smoking to divorce to
employment
—
are in fact
“
contagious
”
across individual groups,
the dynamics of such contagion are of immense interest to social
scientists. The field of cultural evolution has been modeling in-
formation transmission for several decades, using both epidemi-
ological and social network models in their approach (24).
Broadly, social contagion models allow simulating the speed at
which individuals receive information and how past interactions
influence their future behavior (13). These models focus on a
handful of key parameters, which can be grouped as 1) degree
centrality, 2) eigenvector centrality, 3) diffusion centrality, and 4)
betweenness centrality/bridging (18). While one might not wish to
be central in an HIV infection network, centrality is viewed as an
advantage in most social networks and is correlated with financial
success (25) and well-being (26). Degree centrality captures
“
popularity,
”
the sheer number of connections an individual
might have, and the speed at which these individuals can easily
transmit information to a wide group at once. Eigenvector cen-
trality, which captures how many well-connected others one is
connected to, has been used to study social status and scape-
goats (27). Diffusion centrality is a measure of
“
reach,
”
showing
how well-positioned an individual is to spread and hear about
information. Finally, betweenness centrality, or bridging, captures
“
social chameleons
”
who connect otherwise disparate groups.
Interestingly, all of these positions appear to be context general: If
an individual is central in one network, they are likely to be central
in another, and so forth (18).
Each of these four
“
centralities
”
has different disciplinary ori-
gins: the idea of degree centrality began with sociologist and
philosopher Georg Simmel (28); eigenvector centrality is a con-
cept from graph theory, first used by mathematician Edmund
Landau in an 1895 paper on chess tournaments (29); diffusion
centrality became popular in recent literature by economists in-
terested in the speed of information transmission (30); and be-
tweenness centrality, or bridging, comes from sociology literature
analyzing the creation and upkeep of social capital (31). In other
words, the development of these social contagion models was
itself an interdisciplinary enterprise from the beginning.
Since its creation, network analysis has allowed researchers to
apply
new tools while revisiting old questions about social influ-
ence. For example, researchers have investigated the types of
individuals in a network to whom people gravitate, and hence may
be more influential at spreading information of various types (26).
Computational modeling methods have been used to show
quicker consolidation of majority opinion and more successful
spread of initially unpopular beliefs in populations characterized
by greater susceptibility to social influence (32).
Other work using standard economic games has found that
people give less money to those who are more socially distant
(33). This has important implications when combined with the role
that homophily plays in social networks, with many schools being
heavily segregated by race, for example (34). Given the race-
based economic disparity in many countries, this analysis has
taught us that increasing the transfer and exchange of capital
between people of different backgrounds must accompany ef-
forts to interlink their social networks better.
Example 2: The Spread of Infectious Disease.
Sociologists have
been integral to guiding the development of network models,
given how ubiquitously they help explain the spread of anything
from disease to innovation (35). For example, most infectious
diseases spread through human contact, making the study of
Buyalskaya et al.
PNAS
|
3of11
The golden age of social science
https://doi.org/10.1073/pnas.2002923118
Downloaded at California Institute of Technology on January 22, 2021