Transcript Contagion in real social networks

Contagion in real social networks:
insights from social insects
Michael Otterstatter
Zachary Jacobson
What is a social network?
Influence
Disease
Information
Resources
2
Disease spread in social networks
Meyers et al. 2005. J. Theor. Biol.
WHO 2005
3
Problem: disease spread is unobservable
A possible solution: study transmission of
observable proxies for contagious disease
▫ infectious spread of behaviour (behavioural contagion)
4
A novel approach
The transmission of behaviour, as a proxy
for disease, can be studied directly in
social insect networks
Here, we ask
•
does mobility behaviour spread contagiously
among bumble bees via social contact?
•
is the contagious spread of behaviour a useful
proxy for the spread of disease?
5
Materials and methods
• Bumble bees (Bombus spp.)
▫ 7 colonies, reared from wild queens
▫ colonies maintained in the lab under constant light,
temperature
▫ bees allowed to forage at will in flight cage
▫ observations throughout colony cycle (3-20+ bees)
• Automated behavioural tracking
▫ Ethovision software used for 331 hr hive observations,
tracking movement and contacts between nestmates
▫ all observations and analyses are based on the
natural behaviour of bees within their hive
6
Lifecycle of bumble bees
7
Bumble bees in the lab
‘bee-movie’
8
Automated tracking of bee behaviour
5 cm
behavioural tracking
software
video
camera
colony
Example of
movement traces
from a single
colony
flight cage
9
Three analyses of bee mobility behaviour
10
1. Analysis of isolated bees
Do isolated inactive bees
‘activate’ spontaneously
after a fixed interval?
11
2. Analysis of interacting bees
mobility behaviour
z
contact
rates
zz
Are inactive bees ‘activated’ by
contacts from mobile nestmates?
12
3. Analysis of all bees within a hive
In an active hive, is a bee’s
movement behaviour
related to its recent contact
rate with nestmates?
13
Results of bee behaviour analysis
14
1. Mobility behaviour of isolated bees
Isolated bees show no
inherent rhythm of activity
15
2. Mobility behaviour of interacting beesz z z
Inactive bees
…that became active (n=89)
…that remained inactive (n=21)
rec’d 1.46 contacts/min
rec’d 0.14 contacts/min
Inactive bees receiving many contacts from mobile
nestmates tend to become mobile themselves
16
2. Mobility behaviour of interacting beesz z z
After a ‘refractory’
period, contacts
from nestmates
increase a bee’s
probability of
becoming mobile
(logistic regression)
17
3. Mobility behaviour of all nestmates
In most cases, social contacts cause
mobility behaviour to spread between bees
and mobility feeds back to cause increased
contacts (bi-directional causality)
Granger Causality Statistics
Uni-directional causality
Contact causes
Mobility
4 bees
Mobility causes
Contacts
5 bees
Bi-directional
causality
17 bees
No causality
8 bees
(summary results from multivariate time-series analysis)
18
Predicted dynamics of groups
When individuals behave as we observed:
We expect group behaviour like this:
Simulated activity of social group
(Goss & Deneubourg, 1988)
19
Observed dynamics in bee hives
In bee hives, activity level showed stable cycles as predicted
20
Observed dynamics in bee hives
Also, average rates of contact within hives showed stable cycles
21
Spread of behaviour and disease
Importantly, these results suggest that the basic
underlying ‘model’ of behavioural contagion and
disease contagion may be the same:
Behavioural contagion:
Disease contagion (SIR model):
22
Conclusions
• Mobility behaviour spreads contagiously among
bumble bees through social contact
• Social transmission of mobility, like disease,
results in oscillatory dynamics at group level
• Studying observable transmission of behaviour
offers a way to understand the unobservable
spread of disease in social networks
23
Acknowledgements
Technical assistance:
Kieran Samuk, Athena Fung
Funding:
Health Canada Postdoctoral Fellowship Program
24