1. Social Media for Safety
Characterizing Online Interactions
between Citizens and Police
Niharika
Sachdeva
Ponnurangam
Kumaraguru,
Munmun
De
Choudhury
HCI
2016
CERC,
IIITD
and
Georgia
Tech
University,
USA
niharikas@iiitd.ac.in

2. cerc.iiitd.ac.in
Expressing Safety Concerns
2

3. cerc.iiitd.ac.in
Traditional Methods
– Police
rely
on
one
–
way
communicaKon
to
reach
ciKzens
3
69%
45%
hRp://www.accenture.com/us-­‐en/Pages/insight-­‐police-­‐ciKzen-­‐survey-­‐policing-­‐social-­‐media.aspx

4. cerc.iiitd.ac.in
Better Interactive Medium for Citizen
Participation
4
50%
Website
and
Portals
23%
Smartphone
81%
Preferred
Plaorm
hRp://www.accenture.com/us-­‐en/Pages/insight-­‐police-­‐ciKzen-­‐survey-­‐policing-­‐social-­‐media.aspx
Social
Media
may
be
best
channel
for
ciKzen-­‐police
engagement

5. cerc.iiitd.ac.in
Police Use Social Media
5

6. cerc.iiitd.ac.in
Top Officials also on Twitter
6
InteresKng
Hashtags
Strike
a
conversaKon

7. cerc.iiitd.ac.in
#myNYPD
7

8. cerc.iiitd.ac.in
Measuring Human Behavior
- Exploring
the
feasibility
of
social
media
in
quanKfying
aRributes
of
communicaKon
- IdenKfying
behavioral
aRributes
like
affecKve
expression,
engagement
and
social
and
cogniKve
response
processes
8
CiKzen
to
CiKzen
CiKzen
to
Police
Police
to
CiKzen
Police
to
Police

9. cerc.iiitd.ac.in
Research Questions
– RQ
1:
Topical
Characteris3cs
- Nature
of
content
and
topics
that
characterize
social
media
discussion
threads
– RQ
2:
Engagement
Characteris3cs
- How
do
ciKzens
and
police
engage
in
social
media
discussion
threads?
– RQ
3:
Emo3onal
Exchanges
- Nature
of
emoKons
and
affecKve
expression
that
manifest
on
social
media
– RQ
4:
Cogni3ve
and
Social
Orienta3on
- What
are
the
linguisKc
aRributes
that
characterize
cogniKve
and
social
response
processes?
9

10. cerc.iiitd.ac.in
Methodology
10
85
Public
and
official
Police
Department
Average
age
3
years
(from
2010
–
April
2015)
47,474
wall
posts
and
85,408
status
updates

11. cerc.iiitd.ac.in
Data Categorization
11
DT
w/
≥
1
Comment
P&C
C
Total
DT
85,408
47,474
46,845
24,984
5,519
17,196
41,326
7,788
PP&C
CP&C
PC
CC

12. cerc.iiitd.ac.in
Measures of Behavior
12
Topics
• N
Gram
Analysis
• K-­‐means
Clusters
Engagement
• No.
of
police
and
ciKzen
who
comment
in
DTs
• DisKnct
ciKzens
who
comment
in
DTs
• Shannon’s
Wiener
Diversity
index
• Average
no.
of
likes
and
comments
Emo6onal
• Valence
• Arousal
Social
and
cogni6ve
• Interpersonal
Focus
• Social
OrientaKon
• CogniKon
LIWC
and
Anew
DicKonary
LIWC
DicKonary

13. cerc.iiitd.ac.in
Topic Characteristics
–
13
Unigram
Freq.
Unigram
Freq.
rules
0.015
safety
0.012
safety
0.014
following
0.011
violaKons
0.014
noKce
0.010
challans
0.011
prosecuted
0.009
please
0.011
movement
0.008
ciKzens
0.01
complaint
0.008
Focus
on
advisories,
the
status
of
different
cases
being
invesKgated
(U
=
700,
p
<
.05,
z
=
−3.57)

14. cerc.iiitd.ac.in
Topic Characteristics
14
Most
posts
tend
to
request
police
to
take
acKon
on
their
complaints
Unigram
Freq.
Unigram
Freq.
please
0.026
people
0.022
take
0.021
please
0.02
acKon
0.019
one
0.019
people
0.019
take
0.016
one
0.019
acKon
0.015
Kme
0.017
Kme
0.015
near
0.017
number
0.013
Higher
Reference
to
“people”

15. cerc.iiitd.ac.in
Clusters of Topics
– Police
iniKated
discussions
are
more
focused
than
ciKzen
iniKated.
15
Awareness
drive
/
safety
campaigns
Road
sense
is
the
offspring
of
courtesy
and
the
parent
of
safety
Prosecuted
/
ac6on
taken
reports
Ac3on
taken
by
[Withheld],
Reg
your
tweet
pe33on,
@[withheld];
33
parking
tag
&
6
no
parking,
1
foot
path
parking.
Cases
booked
on
hospital
road
Advisories
on
situa6ons
Good
-­‐-­‐
Morning
to
all
the
Commuters
of
Shillong
City,
there
is
heavy
movement
over
NH
-­‐
40
–
44
and
Madanr3ng
down
side,
Lumdiengjri
area
stretch.
Please
do
not
overtake

16. cerc.iiitd.ac.in
Clusters of Topics
16
– Police
iniKated
discussions
are
more
focused
than
ciKzen
iniKated.
Apprecia6on
Hear3est
congratula3ons
to
[withheld]
police
for
nabbing
[withheld]
agent
within
24hrs.
wow!!!
Kudos
and
respect
Newspaper
ar6cles
Please
ACT:
h]p://3mesofindia.india3mes.com/videos/news/…
Ci6zen
6ps
and
complaints
4th
Nov
2014
[withheld]:
Driving
in
wrong
side
at
Teghoria
U
Turn
Neighbourhood
problems
“Learn
from
the
Delhi
incident
and
ensure
that
no
buses
in
Kolkata
have
3nted
glasses.
One
such
bus
was
spo]ed
on
Gariahat
road
Regn.
#.
[Withheld].
Kindly
take
appropriate
ac3on.
Thank
you
Missing
people
“Sir
plz
help
find
my
nephew,
he
is
missing
since
today
morning,
he
is
from
kodagu,
contact
[withheld]

17. cerc.iiitd.ac.in
Research Questions
– RQ
1:
Topical
Characteris3cs
- Nature
of
content
and
topics
that
characterize
social
media
discussion
threads
– RQ
2:
Engagement
Characteris3cs
- How
do
ciKzens
and
police
engage
in
social
media
discussion
threads?
– RQ
3:
Emo3onal
Exchanges
- Nature
of
emoKons
and
affecKve
expression
that
manifest
on
social
media
– RQ
4:
Cogni3ve
and
Social
Orienta3on
- What
are
the
linguisKc
aRributes
that
characterize
cogniKve
and
social
response
processes?
17

18. cerc.iiitd.ac.in
Engagement Characteristics
– Content
GeneraKon
18
Police
+
CiKzens
55,028
1,79,176
17,124
12,630
CiKzens
Only
54,982
1,79,176
17,081
12,630
Entropy
4.39
4.96
3.23
3.6
Police
CiKzen
26%
lower
CP
&C
discussion
threads
might
be
contribuKng
more
towards
social
capital
building
6
(due
to
engaging
both
police
and
ciKzens)
than
those
involving
ciKzen
commentary
only

19. cerc.iiitd.ac.in
Engagement Characteristics
– Content
GeneraKon
19
Police
+
CiKzens
55,028
1,79,176
17,124
12,630
CiKzens
Only
54,982
1,79,176
17,081
12,630
Entropy
4.39
4.96
3.23
3.6
Police
CiKzen
10.28%
lower
Lower
entropy:
large
number
of
comments
are
posted
by
a
small
number
of
ciKzens
and
police
CP&C
Contribute
less
towards
police
endeavours
to
obtain
mass
public
parKcipaKon
in
community
policing

20. cerc.iiitd.ac.in
Engagement Characteristics
20
Comments*
Likes**
Avg.
Std.
dev
Avg.
Std.
dev
Cp&c
3.34
19.19
9.4
253.85
Cc
3.69
13.79
13.38
201.57
– Content
InteracKon
Ci#zen
post:
“My
family
and
I
are
gegng
the
unwanted
calls
from
the
given
number
[withheld].
Especially
he
is
misbehaving
with
a
female
member.
My
Number
is
-­‐
[withheld]”
Police
reply:
“Dear
[withheld],
Please
visit
at
your
nearest
Police
Sta3on
and
lodge
a
complaint
with
details
and
they
will
assist
you
in
this
regard...
Thankyou”
9.49%
lower
29.75%
lower
Police
suggests
an
appropriate
acKon
and
the
discussion
tends
to
close
early,
resulKng
in
lower
interacKon

21. cerc.iiitd.ac.in
Engagement Characteristics
– Content
InteracKon
21
Comments
Likes
Avg.
Std.
dev
Avg.
Std.
dev
PP&C
19.68
86.17
114.71
805.55
PC
9.88
74.92
88.05
1025.53
30.28%
higher
99.19%
higher
PP&C
discussions
are
mainly
advisories
where
police
requests
ciKzens
for
some
acKon
or
shares
informaKon
with
them,
these
may
generate
more
engagement

22. cerc.iiitd.ac.in
Research Questions
– RQ
1:
Topical
Characteris3cs
- Nature
of
content
and
topics
that
characterize
social
media
discussion
threads
– RQ
2:
Engagement
Characteris3cs
- How
do
ciKzens
and
police
engage
in
social
media
discussion
threads?
– RQ
3:
Emo3onal
Exchanges
- Nature
of
emoKons
and
affecKve
expression
that
manifest
on
social
media
– RQ
4:
Cogni3ve
and
Social
Orienta3on
- What
are
the
linguisKc
aRributes
that
characterize
cogniKve
and
social
response
processes?
22

23. cerc.iiitd.ac.in
Emotional Expressions
– NegaKve
senKment
higher
in
ciKzen
iniKated
threads
23
CP&C
CC
Avg
Std.
dev
Avg
Std.
dev
NA
0.021
0.03
0.018
0.04
Anx
0.001
0.01
0.003
0.02
Anger
0.006
0.02
0.005
0.02
Arousal
4.4
1.74
3.9
2.16
16.67%
higher
in
CP&C

24. cerc.iiitd.ac.in
Emotional Expressions
– NegaKve
senKment
higher
in
ciKzen
iniKated
threads
24
Cp&c
Cc
Avg
Std.
dev
Avg
Std.
dev
NA**
0.021
0.03
0.018
0.04
Anx**
0.001
0.01
0.003
0.02
Anger**
0.006
0.02
0.005
0.02
Arousal**
4.4
1.74
3.9
2.16
200%
higher
in
Cc
I
am
just
worried
if
Hyderabad
Traffic
Police
[HTP]
makes
things
worse
like
always
and
create
more
chaos.
Frankly
speaking...
it's
the
lower
income
group
or
the
people
who
are
not
aware
using
high
beams.
Try
to
educate
people
on
road.

25. cerc.iiitd.ac.in
Emotional Expressions
– NegaKve
senKment
higher
in
ciKzen
iniKated
threads
25
CP&C
CC
Avg
Std.
dev
Avg
Std.
dev
NA**
0.021
0.03
0.018
0.04
Anx**
0.001
0.01
0.003
0.02
Anger**
0.006
0.02
0.005
0.02
Arousal**
4.4
1.74
3.9
2.16
12.82%
higher
in
Cpc
Higher
arousal
and
nega#ve
affect
to
be
markers
of
sensi#sa#on
because
of
crime!

26. cerc.iiitd.ac.in
Research Questions
– RQ
1:
Topical
Characteris3cs
- Nature
of
content
and
topics
that
characterize
social
media
discussion
threads
– RQ
2:
Engagement
Characteris3cs
- How
do
ciKzens
and
police
engage
in
social
media
discussion
threads?
– RQ
3:
Emo3onal
Exchanges
- Nature
of
emoKons
and
affecKve
expression
that
manifest
on
social
media
– RQ
4:
Cogni3ve
and
Social
Orienta3on
- What
are
the
linguisKc
aRributes
that
characterize
cogniKve
and
social
response
processes?
26

27. cerc.iiitd.ac.in
Social and Cognitive Orient.
– Discussion
threads
involving
just
the
ciKzens
are
highly
self-­‐aRenKon
focused
27
Likely
ciKzens
mostly
express
their
own
concerns
that
they
face
with
others
CP&C
CC
ppron
0.062
0.059
0.045
0.056
i
0.008
0.017
0.014
0.033
shehe
0.002
0.01
0.003
0.003
they
0.005
0.013
0.008
0.008
75%
More
I
have
lived
in
the
UK
and
all
the
3me
I
have
never
heard
anyone
honking.
Honking
is
not
required
if
you
know
how
to
drive
[...]
Can
anyone
advise
me
where
to
complain
if
I
see
anyone
who
don't
comply
?

28. cerc.iiitd.ac.in
Why it matters?
– Understanding
that
helps
police
improve
policing
and
community
sensing
- Facebook
can
be
used
to
record
and
sense
behavioural
aRributes
such
as
engagement,
emoKons,
and
social
support
– Enable
police
and
ciKzen
community
to
enhance
emoKonal
support
to
residents
experiencing
safety
issues
- Discussion
threads
with
police
and
ciKzen
commentary
showed
reduced
levels
of
anxiety,
showing
police
interacKons
can
be
calming
to
ciKzens.
28

29. cerc.iiitd.ac.in
Technological Implications
– Helping
communiKes
to
make
consensus
based
decisions
regarding
support
and
acKons
they
seek
from
police
– Help
gauge
changing
emoKons
and
behaviour
among
ciKzens
- Timely
and
early
predicKve
analyKcal
systems
– Sense
and
record
the
reacKons
of
ciKzens
and
share
these
records
with
decision
makers
- Take
Kmely
measures
and
gain
beRer
insights
29

30. cerc.iiitd.ac.in
Acknowledgement
– TCS
research
for
funding
the
project
– Members
of
Cybersecurity
EducaKon
and
Research
Centre
(CERC)
and
Precog
who
have
given
us
conKnued
support
throughout
the
project
– Special
thanks
to
Siddhartha
Asthana
30

31. Thank
you!
niharikas@iiitd.ac.in
cerc.iiitd.ac.in