Integration and Automation in Practice: CI/CD in Mule Integration and Automat...
Barcelona presentation reduced
1. Friends in protest: behavioural styles, networks
and affordances of four political groups on
Facebook
Giuseppe A. Veltri (University of Leicester)
Matteo Gagliolo (Universit´e libre de Bruxelles)
EUSN, Barcelona, July 4th, 2014
2. Introduction: FB for research?
Actions don’t lie [Chamley 2004]
Large amounts of data with little effort
Observational: captures actual behaviour (not self-reported)
”Big” data
Social dynamics
Cultural evolution, opinion dynamics
Issues..
Self-censoring
Selection biases
Affordances
3. Behavioural style of minorities/majorities
In Moscovici’s theory (1984) of minority influence, one important
aspect is that different behaviour styles of members that a
minority group has compared to the majority.
Gerard (1985) outlined the behavioural features of minority groups
drawn from both theory and experimental results.
4. Affordances of FB
Ever-newer waters flow on those who step into the same rivers.
[Heraclitus]
Stream moves fast
Echo chambers (edgerank)
Algorithmic gatekeeping
Illusion of visibility: writing on
walls with invisible ink
5. Data
Two minority groups at opposite sides:
No TAV
Casa Pound
“Baseline” comparisons: two majority groups
Partito Democratico (PD)
Popolo delle Libert`a (PdL)
Group Posts (2012) Users
No TAV 2740 38175
Casa Pound 591 17438
PD 1503 21216
PdL 3558 7075
6. Hypotheses and questions
Minority groups:
more than majority (re-)defining reality (anchoring)
recruitment
more than majority self boosting
more than majority more informational social influence
more than majority more self-reference behaviour
Both: impact of affordances?
Two main research directions:
activities of the page (admins)
activities of the users
7. Sample definition
For each public page, you can download the entire stream of posts
by the page admins
by others (writing on the wall (excluded here: only NoTAV and PD
allow it))
For each post: all connections
likes
comments (w. timestamp, likes count)
shares (w. timestamp, likes, comments, shares count)
. . . and all the rest (message, links, tags, pics, . . . )
For each post and connection: its author
id, name, gender, language (if person)
nothing else (no likes, friends, posts, . . . )
The set of all posts and their connections defines our sample of FB users
(unique id’s)
8. Network data extraction
Data from a facebook page can be
represented as a (temporal)
two-mode network
a mode of posts
a mode of users
links connect users to post they
liked
post degree = n. likes on the
post
user degree = n. of likes by the
user
. . . same for comments and shares
9. Page activity
Two main roles of FB pages:
producer of content
relayer of content produced elsewhere
on FB: shared posts
outside FB: external links
11. Page activity: links and informational self-referencing
Minority groups: more links within FB
0.25
0.50
0.75
0.00/1.00
0.25
0.50
0.75
0.00/1.00
0.25
0.50
0.75
0.00/1.00
0.25
0.50
0.75
0.00/1.00
notav casapau
pd pdl
link_type
External
FB link
No link
12. User activity
All activities are reliable to be noticed by friends
Only recent comments on recent posts are visible to group
Activity Visibility Target Cost Impact, perceived* Impact, actual
Like Public (count) Ingroup 1 click ”Count me in” Counter +1
Comment Public (count) Ingroup 1 click + text Participation, debate Counter +1
Share User set Outgroup 2 clicks (+ text) Activism, recruitment +1, Friends
Post User set Ingroup 1 click + (link, text) Proposal Friends
13. User activity: Self-boosting
Minority: more likes [Complementary Cumulative Distribution Function]
0.00
0.25
0.50
0.75
1.00
10 1000
nlikes [log scale]
P(X>nlikes)
page
notav
casapau
pd
pdl
18. Affordances: reaction times
Most activity happens within a few hours from publication [CDF]
0.00
0.25
0.50
0.75
1.00
1s 1m 1h 8h 1D 1W 1M 1Y
relative_time [log scale]
P(X<=relative_time)
page
notav
casapau
pd
pdl
comments relative_time
19. Affordances: reaction times
Most activity happens within a few hours from publication [CDF]
0.00
0.25
0.50
0.75
1.00
1m 1h 8h 1D 1W 1M 1Y
relative_time [log scale]
P(X<=relative_time)
page
notav
casapau
pd
pdl
shares relative_time
20. Conclusions
Minority groups:
(re-)defining reality: more anchoring
self-referencing: more content within FB + more share of shares
within groups
self boosting: more likes
more cohesive? Issue: no data on friendship
Impact of affordances:
costlier activities are less frequent
shares least frequent (low time cost, but perceived as more visible?)
most activity within a few hours
21. Open issues
Sample definition (passer-by vs activist)
Fair comparisons (less posts means longer visibility)
User perception (what do they think they’re doing?)
Offline vs. offline, esp. for No TAV: how do peaks of activity relate
to protest events?
Network analysis proper (REM, tnet)
Longer term:
Questionnaires on FB: app with rights
get insights on user’s perceptions, motivations
get access to private data (friends, likes)
22. Thank you for your questions!
Giuseppe A. Veltri <gv35@le.ac.uk>
Matteo Gagliolo <mgagliol@ulb.ac.be>