2. The
Global Village?
The Internet was based on the promise of
enlarging your social world beyond the
limits of the local village
But has it actually worked?
3. Social Brain Hypothesis
• Predicted group size
for humans is ~150
• “Dunbar’s Number”
Monkeys
Apes
Neocortex volume divided by rest of brain
4. The Natural Size
of Human
Communities?
These all have mean sizes of
100-200
Neolithic villages 6500 BC 150-200
Modern armies (company) 180
Hutterite communities 107
‘Nebraska’ Amish parishes 113
business organisation <200
ideal church congregations <200
Domesday Book villages [1087 AD] 150
C18th English villages 160
GoreTex Inc’s structure 150
Research sub-disciplines 100-200
Small world experiments 134
Hunter-Gatherer communities 148
Xmas card networks 154
225-249
175-199
125-149
75-99
Maximum Network Size
325-349
350-374
275-299
300-324
250-274
200-224
150-174
100-124
25-49
50-74
10000
1000
100
10
0-24
Number of Cases
10
9
8
7
6
5
4
3
2
1
0
1
0 10 20 30
“Reverse”
Small World
Experiments
Hunter-Gatherer
Societies
Xmas Card
Networks
Individual Tribes
5. Human
http://www.youtube.com/watch?v=ApOWWb7Mqdo
Social Groups
These aallll hhaavvee mmeeaann ssiizzeess ooff
110000--220000
Neolithic villages 6500 BC 150-200
Modern armies (company) 180
Hutterite communities 107
‘Nebraska’ Amish parishes 113
business organisation <200
ideal church congregations <200
Doomsday Book villages 150
C18th English villages 160
GoreTex Inc’s structure 150
Research sub-disciplines 100-200
Small world experiments 134
Hunter-Gatherer communities 148
Xmas card networks 154
225-249
175-199
125-149
75-99
Maximum Network Size
325-349
350-374
275-299
300-324
250-274
200-224
150-174
100-124
25-49
50-74
10000
1000
100
10
0-24
It was an advertising stunt!
Number of Cases
10
9
8
7
6
5
4
3
2
1
0
1
0 10 20 30
“Reverse”
Small World
Experiments
Killworth et al (1984)
Hunter-Gatherer
Societies
Dunbar (1993)
Luckily, it’s a hoax….
Individual Tribes
Xmas Card
Networks
Her 152 friends recorded for posterity…..?
Hill & Dunbar (2003)
6. Is Your Online Network Bigger than
~150?
Twitter Email
Gonzalez et al. (2011:PLoS-1) Haeter et al (2012: Phys. Rev.
Letts)
• Network size estimated from reciprocated exchanges
• # edges drops off after ~200
7. Has Facebook Really Widened
Your Social World?
• It seems not….
• Modal number of ‘friends’ on Facebook = 150-250
• You may list 100s of friends, but you only talk to a handful
N » 1 million Facebook users
8. BUT….our friends are NOT all the same!
Our social world is less like this
…..and more like this
9. Intimacy, Frequency and Trust
• Relationship between
frequency of contact
and intimacy
• Trust and obligation
seem to be important
0 1 2 3 4 5 6 7 8 9 10
Emotional Mean Time Since Last Contact (Months)
8
6
4
2
0
LOW Emotional Closeness
HIGH
10. The Fractal Periodicity of
Human Group Sizes
Peak at w=5.4
Peak at w=5.2
Sizes of Hunter-Gatherer
Xmas Card
Database
Social Groupings
Database [N=60]
Scaling ratio = exp(2π/w)
Groupings
= 3.2 and 3.3 Zhou, Sornette, Hill & Dunbar (2005)
Hamilton et al (2007)
11. The Expanding
Circles
Our relationships form a
hierarchically inclusive
series of
circles of increasing size
but
decreasing intensity
[ie quality of relationship]
We know all these layers
exist
…and the military
maintain the sequence far
beyond [to ~50,000]
5
15
50
150
Intensity
EGO
500
1500
12. The Military Model
Modern Army Organisation
USA Australia
[1994] [2010]
The need to solve two conflicting requirements:
Section 10 12
Platoon 30 45
Company 126 168
Battalion 650 775
Brigade/Regiment 4000 3750
Division 12,500 15,000
War of Spanish Succession
[1701-1714]
Maximising
cohesion and the number of boots-on-the-ground
13. Network Structure
on Facebook
• Facebook regional network in
April 2008
• 3M nodes with 23M edges
[useable dataset: 92,300
nodes]
• Density-based clustering:
Optimal cluster structure is
4 layers
• Layer sizes correspond exactly
to those found by Zhou et al.
(2005) in F2F networks
….with a scaling ratio of ~3
….AND an added layer at 1.5
Optimal Cluster #
Support Sympathy Affinity
?? clique group group
Cumulative size: 1.6 5.7 17.6 52.2
Predicted size: (1.5) 5 15 50
Arnaboldi et al. (2012)
14. Network Structure on
Twitter
• 205,000 human Twitter
followers, 200M tweets
• Reciprocated postings
• Optimal # clusters = 4 • Layers have same scaling ratio
[~3) and sizes
virtually
identical to
theoretical
layers
Facebook: 1.6 5.7 17.6 52.2
Theoretical: 1.5 5 15 50
15. The Expanding
Circles
… as they really are
• It turns out, as predicted,
that there really is an inner-inner
layer at 1.5
• …perhaps because girls can
have two intimate relationships
(a best girlfriend PLUS a
boyfriend)
….but boys can only
manage one (a girlfriend
or nothing)?
5
15
50
150
1.5
16. Social Bonding Primate-Style
Primate social bonds
seem to involve two
distinct components:
An emotionally intense
component
[= grooming Þ endorphins]
A cognitive component
[=brain size + cognition]
17. • Best predictor of network size is
orbitofrontal prefrontal cortex volume
• In a fine-grained VBM (voxel) analysis:
best predictor of network size is
ventromedial PFC
• 2 of 7 neuroimaging studies showing
correlations between brain region
volume and network size in humans and macaques
Friendship on
the Brain?
18. Importance of Time
Change in Emotional Closeness Daily contact rates per person
Kin
Friends
0 9 18
months
Friendships decline rapidly in
the absence of contact
19. Time really is a
Network Constraint
• Mobile phone dataset from 11
months [20M users and 9 billion
calls]
• As network size [k] gets larger,
o mean call rate asymptotes at ~200
o call diversity declines after a peak at
k≈15
Total calling is time is limited,
and gets distributed more
thinly
• There is a natural limit to
network size, and it is set [in
part] by how thinly social capital
can be invested
Miritello et al. (2013)
20. Just how consistent are
these patterns?
An 18-month longitudinal study of 30
18-year-olds transitioning to University
….for whom we have complete call + text
records and detailed relationship
questionnaires (at start, mid and end)
Roberts et al (2009), Roberts & Dunbar (2010a,b)
21. Stability of Social Signatures
• Alters ranked by
frequency of calls
• Ranking pattern
remains similar across
all three 6-mnth
windows
DESPITE high turnover in in
membership in successive 6-
month windows
[esp. in first interval as
indicated by low Jaccard
index – indexes similarity]
• 25% to top Alter
48% to top 3 Alters
Saramaki et al. (2014)
22. Stability of
Social
Signatures
• Comparison between 3 intervals
• Individual signatures are significantly more
similar over time [dself] than they are to other
individuals’ signatures [dref]
• Picture is identical using Emotional Closeness,
duration of calls and # texts
Saramaki et al. (2014)
Ego 1
Ego 2
23. Three Ways
We Solved the
Bonding Problem
Modern humans
Archaic humans
-.5 0.0 .5 1.0 1.5 2.0 2.5 3.0 3.5
Millions Years BP
Predicted Grooming Time (%)
50
40
30
20
10
Religion and its rituals
Music and dance
Laughter
a cross-cultural trait
shared with chimpanzees
Australopiths
H. erectus
24. Something in the Way
She Moves….?
• A study carried out in
Brazil with very simple
dance moves
Change in Pain
Threshold
Self-in-Other
Index
25. Conclusions
• Human social networks are constrained by (1) cognition
and (2) time
• The internet has increased the distance over which we
can contact network members….
• BUT it has not increased the size or structure of our
networks
• The real limitations [now] are:
o Lack of face-to-face interaction
o Absence of endorphin-based bonding mechanisms