Presentaci de la Sessi

Poden les xarxes socials predir
els resultats electorals?
Can Social Networks Predict
Electoral Contests?
Frederic Guerrero-Solé
Taula de Nova Recerca
#6N2014
Introduction
Big Data and the All Paradigm
The end of the 3 S: Samples, Statistics and Surveys
Social Networks as providers of ALL
However , Is All in Social Networks, can we ask All
and answer All?
Introduction
The Google Flu Trends
Introduction
The Google Flu Trends
Introduction
What about Politics, and Electoral Contests?
Attempts to predict electoral results using Twitter data
Tumasjan, Sprenger, Sandner, and Welpe (2010)
Number of party mentions can be considered a reflection of the vote share.
However
Jungherr, Jürgens, and Schoen (2011)
Refuted these results. Reported Opacity of method of collecting tweets, the
arbitrariness of the sample and the time window selected, concluding that
mentions cannot be considered a predictor of the election results.
Our Experience
How did you do it, in #7Steps?
@guerrerosolef: using TAGS, looking for samples #step1
How did you do it?
@guerrerosolef: choosing #25N for a neutral discussion #step2
How did you do it?
@guerrerosolef: development of a SNA software to process and
analyze data #step3
How did you do it?
How did you do it?
How did you do it?
How did you do it?
@guerrerosolef: Assumed power law distribution, homophily and
retweeting as a form of endorsement #step4
How did you do it?
@guerrerosolef: Calculated the overlaps between the
communities of retweeters #step5
How did you do it?
@guerrerosolef: Classified users #citizens #politicians
#media and journalists #step6
How did you do it?
@guerrerosolef: Compared our results with those of CIS #step7
How did you do it?
@guerrerosolef: Belief in the statistical signification of
correlations #step7
Also...
@guerrerosolef: We were able to predict
#PoliticalPolarization
#UsersPoliticalLeaning
#JournalistsPerceivedPoliticalLeaning
#7Conclusions
@guerrerosolef: Not All but Limited predictions paradigm. Researchers
have to refine their questions and be creative #conclusion1
Conclusion
@guerrerosolef: Social Networks can provide answers... to good (not
obvious) questions, every sample can lead to new predictions
#conclusion2
Conclusion
@guerrerosolef: samples must fit the purposes of the research
#conclusion3
Conclusion
@guerrerosolef: the path to significant outcomes is full of bias
#conclusion4
Conclusion
@guerrerosolef: Researchers have to find ways to compare results with
those of real world #conclusion5
Conclusion
@guerrerosolef: the evolution of Social Networks and Predictive
Analytics will shape the way we can predict outcomes such as electoral
contests #conclusion6
Conclusion
@guerrerosolef: Capacity to collect, process and analyze data as a new
source for knowledge gap #conclusion7
Thank you!
Frederic Guerrero-Solé
Taula de Nova Recerca
@guerrerosolef