Crowd psychology and crowd safety: From disaster to prevention

Crowd psychology and crowd safety:
From disaster to prevention
Dr John Drury
School of Psychology
University of Sussex
UK
Crowd psychology and crowd safety
… an unlikely pairing?
Traditional ideas about crowd psychology:
• Mindlessness and barbarism
• Loss of self, loss of control
• ‘Mass panic’
Overview
1. How crowds behave in emergencies
2. The social identity approach
3. The role of crowds in disaster prevention
The ‘received wisdom’
‘Most deaths in night-club fires are due to
“crowd panic”, not the fire itself’
A textbook case:
Cocoanut Grove nightclub fire, 1942 (492
people died)
Cocoanut Grove nightclub fire, 1942
(Chertkoff & Kushigian, 1999)
• The emergency exit door was
locked
• Windows were sealed
• Many died from fumes and then a
ball of fire.
• Enquiry: no implication that crowd
over-reaction or selfishness
caused the deaths.
• The management were
prosecuted for manslaughter and
neglect of building laws.
Conceptual problems with ‘mass panic’
Quarantelli (1960)
• How do we know it’s panic?
• ‘Flight’ is a more scientific term for behaviour.
• To judge a response as
irrational/unreasonable requires a frame of
reference.
Empirical problems with ‘mass panic’
Case studies note a lack of ‘mass panic’:
•
Atomic bombing of Japan during World
War II (Janis, 1951)
•
King’s Cross Underground fire of
1987
(Donald & Canter, 1990)
•
9/11 World Trade Center disaster (Blake,
Galea, Westeng, & Dixon, 2004)
Reviews of the evidence also note the
absence of ‘panic’
• Panic is actually rare in crowds (Brown, 1965;
Johnson, 1988; Keating, 1982)
But what does this mean?
Generic or common reactions?
• Loss of behavioural control
• Selfishness
• Disorderly responses
BUT
• Cooperation is relatively common - within and
across emergencies
• Evacuations are often orderly
Explanations
1. Social norms
(Johnson, 1987, 1988)
• Explaining or redescribing?
The Beverly Hills Supper Club
fire, 1977 – 156 fatalities
(Johnson, 1988)
Police records: interviews with
630 people present
The elderly were helped more
than the less vulnerable
Explanations
2. Ties of affiliation
(Mawson, 2005, 2007)
Fire at the Summerland leisure complex, 1973 – 50 fatalities
(Sime, 1983)
•
People tried to exit in small (family) groups, not alone
•
People prefer to stay with loved ones, and even die together,
rather than escape alone.
• How do strangers behave with each other
in an emergency?
7th July 2005 London bombings
• 56 people died
• 700+ injuries
Emergency services
didn’t reach all
the survivors
immediately
The crowd
was left in the dark for
up to 20 minutes or more
The study
• Accounts from over 146 witnesses
• 90 of whom were survivors
• 17 interviewed/written responses
Drury, Cocking, & Reicher (2009) International Journal of Mass Emergencies and Disasters
‘Helping’ (versus personal ‘selfishness’)
(Helping = giving reassurance, sharing water, pulling people
from the wreckage, supporting people up as they
evacuated, tying tourniquets)
‘I helped’
‘I was helped’
‘I saw help’
‘Selfish’ behaviours
Contemporaneous
newspaper
accounts
Archive
personal
accounts
Primary data:
Interviews and
e-mails
57
17
140
3
42
29
50
11
13
10
17+
4
Drury, Cocking, & Reicher (2009) International Journal of Mass Emergencies and Disasters
‘I remember walking towards the stairs and at
the top of the stairs there was a guy coming
from the other direction. I remember him
kind of gesturing; kind of politely that I
should go in front- ‘you first’ that. And I was
struck I thought, God even in a situation
like this someone has kind of got manners,
really.’
(LB 11)
Drury, Cocking, & Reicher (2009)
International Journal of Mass Emergencies and Disasters
Why did people help each other?
Possibility of death
Not going to die
With affiliates
With strangers
Archive
personal
accounts
Interviews
and
e-mails
68
2
8
57
12
1
2
15
Drury, Cocking, & Reicher (2009) International Journal of Mass Emergencies and Disasters
‘SOCIAL IDENTITY’
• We all have personal identities (‘me’)
• We also each have multiple social identities,
based on group memberships (‘us’)
• Physical crowds can become united through
shifting from personal to shared social identity
(Reicher, 1984)
Applying social identity to emergency
crowds
Hypothesis 1: Antecedence of shared social identity
– An emergency can create a sense of ‘common fate’
(new group boundaries)
– People use this ‘common fate’ to define their identity
Hypothesis 2: Consequences of a shared social identity
– ‘They’ = ‘us’
– Caring about others
– Giving social support
Int: “Comparing to before the blast happened what do you think the
unity was like before?”
LB 1: “I’d say very low- three out of ten, I mean you don’t really think
about unity in a normal train journey, it just doesn’t happen you
just want to get from A to B, get a seat maybe”
(LB 1)
‘Me’ in relation to other individuals
Interviewees’ references
to ‘we-ness’:
•
•
•
•
•
‘unity’,
‘together’
‘similarity’
‘affinity’
‘part of a group’
• ‘everybody, didn’t matter
what colour or nationality’
• ‘you thought these people
knew each other’
• ‘teamness’[sic]
• ‘warmness’
• ‘vague solidity’
• ‘empathy’
Afterwards: ‘Us’ in relation to the bombing
Accounting for help
Archive
personal
accounts
Interviews
and
e-mails
Shared fate
11
5
Unity
20
11
Disunity
0
1
Drury, Cocking, & Reicher (2009) International Journal of Mass Emergencies and Disasters
Int: “Can you say how much unity there was on a
scale of one to ten [after the explosion]?”
LB 1: “I’d say it was very high I’d say it was seven
or eight out of ten.”
(LB 1)
• Almost all who referred to ‘unity’ also referred
to ‘help’
Drury, Cocking, & Reicher (2009) International Journal of Mass Emergencies and Disasters
Further research aims
1. Test the social identity account with a
representative sample
Further research aims
1. Test the social identity account with a
representative sample
2. Examine the (mediating) role of expected social
support
Further research aims
1. Test the social identity account with a
representative sample
2. Examine the (mediating) role of expected social
support
3. Examine relationship between social identity and
observing the (solidarity) behaviour of others
Our opportunity:
• The 2012 Solidarity Survey (MIDE UC
Measurement Center at Pontificia Universidad
Católica de Chile and Hogar de Cristo)
• We piggy-backed that survey to ask about the
Chile earthquake of 2010
Drury, Brown, Gonzalez, & Miranda
(2015) European Journal of Social
Psychology
Chile earthquake and tsunami,
2010
• Earthquake occurred off the coast of
central Chile on 27th February 2010
• Triggered a tsunami in coastal towns
• 521 died, 9% of the population in the
affected areas lost their homes
• Emergency services overwhelmed
• Solidarity was therefore essential for
survival and recovery
Drury, Brown, Gonzalez, & Miranda (2015) European Journal of Social Psychology
Questionnaire
Survey of a representative sample of 1240 adults
in the affected regions (non-affected regions
excluded)
Results 1: Effects of social identification
Give support
Common
fate
Social
identification
Coordinate
Expect
support
Results 2: Effects of others’ behaviour
Others’
solidarity
Give support
Common
fate
Social
identification
Coordinate
Expect
support
Results 3: Moderation
Others’
solidarity
Give support
Common
fate
Social
identification
Coordinate
Expect
support
The story so far …
1. Cooperation among strangers is common in
emergencies
2. Social identity provides:
– Motivations to provide social support
– Expectations of social support
2. From disaster to prevention
The public as the ‘fourth emergency
service’
– Necessary because of absence of professional
responders
• What about potentially dangerous crowd
events?
– Here too professionals may be limited in efficacy
A ‘near disaster’
• 250,000 people (60,000 expected)
• Emergency services overwhelmed
• Exit routes blocked
‘Panic at DJ Fatboy Slim’s Beach party’
‘seaside resort [brought] to the brink
of disaster’ (Guardian, 21st July
2002)
A ‘near disaster’
1. Some people climbed up the lighting rigs
2. Part of the crowd was close to the waterline
as the tide came in, a crowd
tried to evacuate
surge as people
3. Risk of crushing; some participants became
distressed
BUT it WASN’T the disaster that people feared!
Research questions
1. To what extent did social identity processes
in the crowd explain resilient outcomes (e.g.,
safety)?
2. How did the professional groups and crowd
participants perceive that disaster was
averted?
(Intergroup perspective)
Drury, Novelli, & Stott (2015) European Journal of Social Psychology
Methods
• Questionnaire survey
– (n = 48)
• Interviews
– Crowd participants (n = 10)
– Professional groups (n = 10)
• Contemporaneous archive data
–
–
–
–
–
Video
News reports
Message-board material
Official materials
Written accounts by participants
Drury, Novelli, & Stott (2015) European Journal of Social Psychology
Survey results
b = .54,
p = .003
Crowd
identification
Expecting others
to help
Direct effect, b = .23, p = .18
Indirect effect, b = .32, 95% CI [0.09,
0.69]
Drury, Novelli, & Stott (2015) European Journal of Social Psychology
b = .60,
p < .001
Feeling
safe
Survey results
b = .71,
p < .001
Crowd
identification
Trusting others in
an emergency
Direct effect, b = .17, p = .37
Indirect effect, b = .38, 95% CI [0.16,
0.74]
b = .54,
p = .001
Feeling
safe
Drury, Novelli, & Stott (2015) European Journal of Social Psychology
How was disaster averted?
Police perspective
1. Own organizational resilience:
‘Dunkirk Spirit’ [ ]
Individual Acts of near-heroism on the street/beach (esp.
where crushing occurred)
(Minutes from police debrief)
2. Control, coercion…
Drury, Novelli, & Stott (2015) European Journal of Social Psychology
How was disaster averted?
Steward perspective
Crowd perspective
3. Crowd self-organization:
I’m absolutely of the opinion that it was the crowd that
stopped the disaster.. none of the barriers, none of the
coppers, none of the stew.. stewards, none of the
alleged things that were put into place .. to protect the
crowd I don’t think any of that mattered, I think it was
the crowd that kept everything together
(Crowd participant 4)
Drury, Novelli, & Stott (2015) European Journal of Social Psychology
‘Working with’ the crowd – consent and
coordination
Coordinated and careful evacuation:
...the key to the event, the key to the the fact that it didn't go
wrong is only really because the crowd allowed it not to go
wrong, they were happy, they were content, they were
informed and the mood was great
(Safety advisory group member, Big Beach Boutique, 2002)
Drury, Novelli, & Stott (2015) European Journal of Social Psychology
‘Working with’ the crowd – defining norms
How are we going to get him down? So I said, you ask the voice of
God [DJ Fatboy Slim], turn the music down, ask the voice of God
very nicely to say [ ] ‘please get down, because the party can’t carry
on until you’re back on the ground, but do it safely please.’ Peer
pressure will bring him down, and he won’t get his head kicked in,
and that will stop anybody else climbing up. Music came down,
voice of God came over, he waved a bit, everyone cheered, and
they’re all going ‘down, down, down’, so he comes sliding all the
way down, everyone cheers and that’s it. No one else climbed a
lamp post all night.
(PG8 – city council environmental health officer)
Drury, Novelli, & Stott (2015) European Journal of Social Psychology
Hajj
Mount Arafat
Hajj
• 2006: 346 pilgrims died as they attempted to
“stone the devil” at Jamarat Bridge
– Official explanations: ‘unruly pilgrims’
– Crowd ‘physics’ explanation: flow and density
2015: 2000+ pilgrims died outside Mecca at a
cross-roads
Hajj 2012: A safe event
Holy Mosque: 356,800m2 (88.2 acres)
Total capacity: two million*
Average crowd density level of at least four people per square metre (4ppm2).
However, at certain locations, levels of density as people get closer to the
Kaaba = 6-8ppm2
Hajj
• How does an event which routinely reaches
dangerous levels density pass without crushing
incidents?
• Limited
psychological
research –
Mecca closed to
non-Muslims
The study
• 1,194 pilgrims
• Languages groups: 420 (35%) were
Arabic speakers, 150 (13%) of Malay,
150 (13%) of Urdu, 120 (10%) of
French, 120 (10%) of Persian, 120
(10%) of Turkish, and 114 (9%) of
English.
• Sampled across the three phases of
the Hajj in which the Holy Mosque
would be most busy.
Alnabulsi & Drury (2014) Proceedings of the National Academy of Sciences
Measures
• Density – people
per square metre
(researchers’
estimate)
• Safety (perceived)
• Identification with the crowd
• Perceived/expected social support
• Management competence
Alnabulsi & Drury (2014) Proceedings of the National Academy of Sciences
Predictors
of safety
• Identification with the crowd (β = 0.22, P < 0.001)
• Perceived support (β = 0.16, P < 0.001)
• Management competence (β = 0.132, P < 0.001)
• As density increased so safety decreased, β = −0.061, P = 0.037.
Alnabulsi & Drury (2014) Proceedings of the National Academy of Sciences
Results 1: moderation
Alnabulsi & Drury (2014) Proceedings of the National Academy of Sciences
Results 2: mediation
Perceived support
b = .56,
p < .001
b = .23,
p < .001
Identification with the
crowd
Safety
Direct effect, b = .41, p < .001
Indirect effect, b = .15, 95% CI
[0.10, 0.20]
Alnabulsi & Drury (2014) Proceedings of the National Academy of Sciences
Warnings
• Density (over 5ppm2) IS dangerous – shock waves
• The argument is not that people never over-react to
threats, but that crowds are not inherently the basis of
over-reaction
• Not all crowds in emergencies are characterized by social
support
– Some crowds are divided; some crowds remain ‘physical
crowds’
– Social identity: the conditions for social support, not
generic behaviours
Conclusions
• Social identification is the basis of:
– Collective resilience in emergencies
– The social support and coordination needed to
prevent disaster in potentially dangerous events
• This analysis has implications for:
– Theories of crowd behaviour
– How we think about crowd safety management
• Community Resilience: the ‘community of circumstance’
Acknowledgements
Colleagues:
Hani Alnabulsi, Rupert Brown, Chris Cocking, Roberto Gonzalez, Daniel Miranda,
David Novelli, Steve Reicher, Clifford Stott
Funders
• Pontificia Universidad Católica de Chile and the Interdisciplinary Center for
Social Conflict and Cohesion Studies, CONICYT/FONDAP/15130009
• Leverhulme Trust, F/00 230/AO
• ESRC, RES-000-23-0446
Contact details
[email protected]
Crowds and Identities research group:
http://www.sussex.ac.uk/psychology/crowdsidentities/
@DrJohnDrury