Connection matters: social networks, stability and health in rhesus macaques


So our last speaker for
this session is Dr. Brenda McGowan, who is a professor
of population health and reproduction here at the UC Davis School
of Veterinary Medicine, she is also the director of the animal behavior
laboratory for welfare and conservation. Brenda’s work focuses on
behavioral biology and ecology, with a real
emphasis on both basic and applied research aspects related to
animal behavior and communication. All with the goal of enhancing wildlife
health, as well as conversation. She’s also at the front edge of research
trying to bring network methods into analyzing complex social
phenomenon in primatology, as well as developing new method,
new, new network based methods. Which are really exciting, which she’s
gonna be talking to us about today, in her talk Connections Matter. So thank you Brenda. [APPLAUSE]
>>Thank you Megan thank you for inviting me this week here today. This is a very exciting
conference to be a part of. So my, my original title had a subtitle
of Social Network Stability and Health in Rhesus Macaques. But I decided that I should probably hone this down to something a little
bit more related to decision making. And so I’m gonna focus my attention on
the, the work that we’ve done under that broader category on group cohesion and
collapse in rhesus macaque societies. And I’m gonna be talking about a series
of studies that my research team and I have done over the last several years, looking at the stability of social groups,
of, of large captive social groups at the California National
Primate Research Center. And we’ve been doing this work in part
to be able to develop new methods for detecting instabilities in social groups
so that we might increase the welfare and health, the health, welfare and
management of these non-human primates. But we’re also interested in the,
the drivers and the mechanisms that affe,
that affects stability in social groups, such that we can look at a,
again at being able to identify detect and prevent ss, ss, instabilities in
a number of different kinds of systems. Biological systems, other social systems,
even societal systems. And to, to,
to sort of emphasize this point, we can look at two very powerful inst,
insta, instability systems.. Systems th, that had instabilities. If we look at the polarization,
for example, of the U.S. House of Representatives over time. From 1949 to 2011,
we see that there’s a collapse in this structure of bipartisan voting,
on across, parties. And it’s, we want to understand the processes
by which these kinds of things occur. And of course, we all remember the 2008,
global financial collapse. But again, if we can determine w, ways in
which that we can identify these kinds of problems, we can go ahead and
prevent them. So my team and I have developed s,
sort of a, a conceptual framework to, to begin to understand how we can look at
processes of stability in social systems. And we, I, we think of it as the drivers
including things like in our non-human primate model demographics such as age,
si, size and sex and size and you know sort of group composition can
ship included in that as well, but also can ship dynamics. How that matrilinear structure is
actually is actually formed and also how much of a, a unrelated membership
is contributing to these dynamics. And also the collection
of individual attributes, which includes past experience such
as rearing, current experience and something that we call
biobehavioral organization, which includes things like personality and
other aspects of individual differences. So if we think of those as the drivers
that affect different kinds of networks or the structures of social systems we
can begin to look at these networks in relationship to how key they are in a
particular in a particular social system, as well as how different networks
correspond to one another. And in a non-human primate example. We are talking we’ll be talking
about status signaling networks and its relationship to aggression networks. And then we also are interested in this
notion of certainty in networks and in our reases mckek model we are talking
about dominance probability. So all of these in term the structural
aspects of of a social systems drive whether or not Robust mechanisms
such as conflict intervention, policing, or conflict resolution
can operate to stabilize systems. And so we’re interested in
this entire system as a whole. So why do we rhesus macaques in, in, for,
for stud, for looking at these kinds of, of, issues and looking at the processes
underlining stability and social systems? Well it turns out that rhesus
macaques are a lot like humans. Um,they are a weed species like we are. They live throughout have a wide
distribution throughout South and Southeast Asia in fact, caused a lot
of problems with respect to human, human-monkey conflict in a number
of those, of those countries. They have a very complex social structure. They’re comprised of,
of large multi-mammal female groups that, where females are philanthropic and
the males disburse. The females inherit there
rank from their mother. And there’s quite a clear dominance
hierarchy in rhesus macaque society. And the California National
Primate Research Center has 24 enclosures of, of groups. Anywhere from 80 to 200
animals per enclosure, in these very complex social
systems that we’re able to study. And, they live in these
half acre enclosures. And we have natural variability
in the stability of these groups. So some of these groups are stable and
some of these groups are not stable. So here’s an example of what
our field enclosures look like. Of, these half acre field enclosures
with all of these different animals and the kinds of structures that they have,
in those enclosures. So for the groups that I’m gonna to be
talking about today we actually studied two different kinds of groups observing
them slightly differently from one from, one to each other. In group A we have, three groups which
we studied for a long period of time. For over a year, or
close to a year in, of study. And one of those groups
actually socially collapsed. In April 2011 after a short
observation period. So we have that kind of collapse. And when I’m, when I say a social collapse
I mean a, what we call a cage war. We literally have to disband
the entire cage and, and, and, and and
reform a new cage as a result of that. The second set of groups we studied,
for shorter periods of time, and we have, a couple of
different groups there where we, after, during the time that we were
observing them we, we saw social class. So this is the kinds of groups that we,
observe. When we observe these groups, we’re looking at multiple
different kinds of networks, so we have many different kinds of behavioral
networks that we’re looking at. Including aggression, and grooming. Subordination networks, and
conflict intervention networks, among others, alliance networks. And we collect these data,
with observers, cageside, to collect the interactions between individuals
in order to develop these networks. And what I’m gonna focus on
today due to lack of time, instead of the entire picture, is I’m
just gonna talk a little bit about what we’ve done looking at the,
our keystone networks. Which our status stil,
status signaling networks, as well as looking at our, the
correspondence between multiple networks, and in this particular case two networks,
aggression and status signalling. So let’s first start with talking
about ours, keystone network, the status signalling network, so you have
an idea of what this is comprised of. So macaques produce, behavior called
the silent bared teeth display. And use it as a signal which is
homologous to the human smile as a, as a means of, displaying subordination
to others in the social group, and, what’s interesting about these,
these silent-bared, silent-bared-teeth space or SBTs is
that there are many different types. And one, is a type that is
produced in peaceful context, which is a formal signal of subordination. Indicating that another
animal is dominant to you. And what’s also interesting about these, these SBTs, is that these animals seem
to have a consensus of subordination. That is, there’s a collective, almost
a collective decision making process, almost kind of analogous to
voting going on with these SBTs. And we’ll talk a little
bit more about that soon. I want to emphasize that
sort of that consensus idea, because one thing we see with these
SBT networks are that there are no circular relationships with them,
they are completely transitive, that is they only go in one direction,
up and up and it is related to dominance. So you only see one,
you never see one individual that gives, an SBT to another individual, that,
that same individual giving an SBT back. And so, it’s a very structured, highly structured hierarchical complex
network among these individuals. And here’s an example of
a stable SBT keystone network. In one of our groups what you’ll note
here is that it’s very hierarchical, it’s very complex in it’s connections,
it’s multi-layered. And it has a number of
indirect paths as well. So I’m just giving you some
examples of some of these stable, SBT networks that we see in our groups. And what we see,
what happens in unstable groups, again these get right before a collapse, as I described earlier, is that we
see that this structure, disappears. So we see that the hiarch, hiarchal
structure of the unstable group in 2011, same group, when it was stable
in 2009 completely disappears. We have fewer nodes that are involved in
the SBT network and we have fewer indir, a fewer indirect paths that
are connecting these individuals. As an example of another group, that show,
wa was unstable, we see the si, similar kind of pattern. So it turns out that,
that the, these SBT networks. It’s the hierarchical complexly connected,
multi, multi-teared, indirect pathways is extremely important to stability
in these, in these In this species. So in summary, with respect to the SBT
network and looking at our keystone network, stable groups tend to
show fully connected SBT networks that have multi-level structure and
more long indirect, signaling pathways. Where as our unstable groups
show less connectivity and SBT networks, loss of nodes, multi,
and what multi-level structure, and a loss of those indirect
signaling pathways. And we’re currently be beginning to
quantify this so that we can look at you know, sort of significant changes not
just qualitative changes in networks. Okay.
So, what about looking at multiple networks and
their correspondence between them? We are, like, we are developed a new, network
technique called joint network modeling. That allows us to model
the interrelationship between two or more networks. Currently we are modeling two networks,
but we’re working on modeling
three networks simultaneously. And in the, in this case, we’re gonna
be talking about these two networks, the whoops. The aggression. I’m sorry, something happened. [LAUGH] There we go. The the aggression network and
our status networks and how they actually
correspond to each other. The way that we do this is we have you,
you know, behaviors can be, are directed and can be directed
to one to the other or vice versa. And so in aggression you can either
have animal A aggressing animal B or animal B aggressing animal A. Or you can have animal a giving status
signaling, status signals to animal b, or the other way around. So we have basically four opportunities
for a directed relationship between these, these individuals. You can, in the simplest case what
we have is, is is no interaction. That you have nothing moving
between two individuals. And then we had nine other possible
scenarios in which you can either have. Only a, a bi-directional aggression,
or you can have bi, at the bottom here, you can bi-directional
aggression and bi-directional signaling, just to give you, just to give you a, an example, this,
this represents the, the, the, the four letter code for, the different
kinds of interactions that you can have. So you basically have four, four- I’m sorry- ten different types of
interactions you can have between dyads. And what we can do is take those ten
interaction, interaction, between dyads and we can look at the number of dyads
that we would expect to show those interactions based upon the probability
of the behavior to begin with, right? So this, we can get an expected value
that these two networks are completely independent of each other. And then we can compare that, to what we
see actually occurring in these groups. So we, we, and what we can get is
a chi-square value that tells us how much how often you expect to see this in,
I’m sorry, how often you observe something in relation to how often you expect,
you are expected to see it. And that can be either being more, eh,
being more expected or less expected. And if you look at the, the column that
you see of the chi-square, you see there’s a lot of variation, some of which is very
small in terms of the chi-square which means that you’re meeting expectations and
some that are very high. Which means that the expectation
is not being met and if you look at this particular,
comparison of two networks in a group, you find that these two
networks are not independent. Okay. So once we have that kind of information, what we can do is we can
begin to tweak the model. The null model being inter, independent. And we can p, p, p, provide b, b, begin to put constraints onto this
model in order to see where the, where, which aspect of the interaction
are important to these animals. So we can take a function one for example. So we first have the just the chi
squared value that is the null model, the independence and then we can
take off a constraint function and like for example, that two way aggression
is really rare and we can see how the high square value changes, as a result of that
and if the chi squared value changes or dives in this particular case, the more
important that particular function is. Okay? So we can do that for
all of these different functions. Function one,
being two way aggressive is rare. Function two,
being two way status signaling is rare. And function three, is that aggression and status are in opposite
directions of each other. So what happens when we
look at this kind of, model with respect to stable and
unstable groups. Well, what we see is that
in stable groups again this the total chi square on the y axis and
the constraint functions on the x axis, we can see that in the stable groups over,
looking at various time points of in, instability is that we see
the stable groups show that, that that aggression that,
that bidirectional regression is rare, that status signalling bidirectional
status signaling is rare. But that the, the relationship between
the opposite pathways of aggression and status is actually quite important. Okay?
So that’s what that is showing. And we see that in all
the stable social groups. Well, what do we see in
the unstable social groups? We see that, that,
that relationship changed. First we see an increase in
the importance of bidirectional, bi directional aggression. And but, and we also, so
that’s what we see in these groups. This group actually recovered,
temporarily and then shi, also, fell apart again in,
in, during this unstable period. And remember we also see,
a discu, a decoupling of the, aggression status signaling
opposite interaction. So, those are just sort of two
different modes in which we can see the instabilities occurring. So in summary stable groups show many more aggression status dyads than
expected in the opposite direction. And as much bidirectional
aggression as expected. But in the unstable groups we see
more bidirectional aggression and we see a loss of this aggression
status interdependence. So, in summary, we think that we’ve found a couple of different
ways in which we can identify and perhaps use this information to predict
when social collapse is gonna occur. At least in our macaque social groups. And these kinds of concepts might be
able to be used in other, other systems. Identification of keystone networks that
are extremely important to the stability of a social system. In our case is the SBT
network in Macaque Society. They are used to maintain
stable social groups and maybe used to anticipate social collapse,
I’ve just said. Because, because we can look at this
hierarchical intransitive structure and the indirect pathways. And we can also identify key social roles
that individuals play as a result of, of looking at these kinds
of networks because it, it, indeed it is the conflict interveners that
receive most of these kinds of signals. In these networks and so we can identity
those individuals to make sure that we have the right kind of individuals in,
in the, in the appropriate positions. And joint network modeling is
a very is a wonderful approach, because a powerful approach, because it’s
applicable to detecting instability in just about any kind of complex
system that has multiple networks. For example we already talked
about social systems and behavioral networks that could be used to
look at ecosystem networks and of course, financial system banking networks,
as we suggest in one of our publications. And we can track the interconnections
across networks to track stability and or to maybe intervene when these
kind of systems are becoming, or are starting to collapse. So in summary,
we think that these are useful models and that there are nine primate model or to
develop the kind of tools to identify and predict and prevent social system, prevent
system, systemic collapse in a variety of different kinds of human and non-human
systems and perhaps to be used to model the kinds of systems that we talked
earlier, that come from human society. And with that I will,
just acknowledge the several people that are involved in the research
including our, our, wonderful ob, observation team, our colleagues,
and our funding sources. Thank you.
[APPLAUSE]

Leave a Reply

Your email address will not be published. Required fields are marked *