1
Supplementary Information
(SI)
Social media enables people
-
centric climate action in
the
hard
-
to
-
decarbonise
building
sector
Ramit Debnath, Ronita Bardhan, Darshil U. Shah, Kamiar Mohaddes
, Michael H. Ramage, R. Michael
Alvarez, Benjamin K. Sovacool
Section 1.
R
esults
Table A1 shows the theoretical framework on Twitter causality discourse that informs our time
-
series
interpretation of high
-
level policy events and Twitter engagement. We us
ed this framework to
qualitatively support our data
-
driven analysis and improve the generalisability of the findings.
Table A1 Causality discourses of climate communication on Twitter, as per Bergez & Al
-
Safaq (2020)
Extreme
-
event factor:
the
mediated highlighting of extreme weather, in a general sense or with a focus
on previous or ongoing events.
Media
-
driven science
communication:
media information mentioning ‘causality discourse’ deriving from parallel media
channels/platforms, be they traditional news channels, newspapers, or broadcast
-
radio news, but also political campaigns, blogs, think tanks, organisations, etc.
Digital
-
action factor:
individual users’ networked generation of ‘causality discourse.’
The
table A2 presents key network characteristics of N1 to N4 that supports the network theory
-
driven
findings in the manuscript. It also
provides
extended data for Figure 3 in the main manuscript.
Table A
2
Macro
-
network characteristics
Network
Year
Density
Avg. degree
Modularity
Avg. clustering
coefficient
N1
2009
–
2012
0.015
2.988
0.
444
0.825
N2
2013
–
2016
0.005
3.447
0.
479
0.784
N3
2017
–
2020
0.002
8.516
0.
667
0.838
N4
2021
0.002
7.919
0.
674
0.846
F
ig A1 illustrates a time
-
series 6
-
month moving averages of embedded emotions in the tweets
as a
function of the public reactiveness to high
-
level climate policy events. It’s extended version is
presented in Figure 2 in the main manuscript with correspondin
g tweets having highest emotional
scores. Fig A1 shows emotions as a ratio of emotional shifts in the public reactions.
2
Fig A1: 6
-
month moving average share of emotions across the 13
-
year time scale.
Fig
ures
A2
, A3 and A4 qualitatively
shows the strength of hashtags in its specific network (N1 and
N2). The size of the word corresponds to its centrality scores in the network, i.e.,
the larger the size of
the word the greater
is its influence in the online discourse. We stress on the mid
-
range eigenvector
scores as
they
denotes the most dynamic hashtags in our data corpus over the 13
-
year period, while
the hashtags with high centrality scores subjectively remain same through the N1 to N4 network.
Fig A
2
: Hashtags for N1 and N2 with ei
genvector centrality scores in the range of 0.1 to 0.3.
N
e
t
w
o
r
k
1
(
N
1
)
N
e
t
w
o
r
k
2
(
N
2
)
3
Fig A
3
: Hashtags for N3 with eigenvector centrality scores in the range of 0.1 to 0.3.
Fig A
4
: Hashtags for N4 with eigenvector centrality scores in the range of 0.1 to 0.3.
N
e
t
w
o
r
k
3
(
N
3
)
N
e
t
w
o
r
k
4
(
N
4
)