In an effort to curb air pollution and cope with congestion, the city of Delhi (India), known to be one of the most populated, polluted and congested cities in the world, has implemented the first phase of #OddEven experiment between January 1st and 15th 2016. During the experiment, vehicles were allowed to move on alternate days based on whether their plate numbers end with odd or even digits. While the local government of Delhi represented by A. Kejriwal (leader of AAP party) advocated for the benefits of the experiment, the national government of India, represented by N. Modi (leader of BJP) strove to demonstrate the inefficiency of such initiative. This particular configuration has led to a strong polarization of public opinion towards #OddEven initiative which provided the scientific community with a unique opportunity to study the impact of political leaning on humans' perception in a large-scale real-world experiment.
We collect data about pollution (US embassy station) and traffic congestion (Google Traffic API) to measure the real effectiveness of the experiment. We use Twitter to capture the public discourse about #OddEven and study the underlying opinion and sentiment based on different dimensions: time, location, topics. Our results revealed a strong influence of political affiliations on the way people perceived the success of the experiment. For instance, AAP supporters were significantly more enthusiastic about #OddEven compared to BJP supporters. However, when we limit the analysis to only people who experienced #OddEven (i.e., living inside Delhi), the differences in opinion fade away.
To cite this work:
Tahar Zanouda, Sofiane Abbar, Laure Berti-Equille, Kushal Shah, Abdelkader Baggag, Sanjay Chawla, Jaideep Srivastava. "On the Role of Political Affiliation in Human Perception: The Case of Delhi OddEven Experiment" In proceedings of the 9th International Conference on Social Informatics (SocInfo 2017). Oxford, UK.
1. “OddEven” experiment through the lens of
social media
SofianAbbar,TaharZanouda,LaureBerti-Equille,KushalShah,Abdelkader
Baggag,SanjayChawla,JaideepSrivastava
Doha,10th August2016.
2. Context of Study
46 millions
people [1]
The most polluted
city worldwide[2]
Congested
City
[1] Population of National Capital Region http://goo.gl/WUSOkI
[2] World Health Organization Urban Ambient air pollution database 2016 http://goo.gl/03emdT
01
4. “OddEven” Experiment
• An initiative led by AAP
leader, Arvind Kejriwal.
• Allow vehicles to move on
alternate days based on the
registration numbers
ending with odd or even
digits for two weeks.
03
5. Research Question: What is the impact of political
affiliation on people’s perception about OddEven? 04
6. Experiment Timeline
The first phase of the 15-day
pilot took place at Delhi from
January 1st to January 15th
2016.
Beforeexperiment After experimentDuring Experiment
After the
announcement,
people were torn
between support
and opposition
After the
experiment, people
were questioning
the success of the
initiative.
05
FirstWeek
AfterExperiment
SecondWeek
AfterExperiment
FirstWeek
DuringExperiment
SecondWeek
DuringExperiment
FirstWeek
BeforeExperiment
SecondWeek
BeforeExperiment
8. Data: Types & Sources
07
Traffic Air QualityPublic OpinionData Types
Google Traffic API DPCC ProgramTwitter
Data
Sources
9. Posted by users .
We extracted their tweets, profiles, friends
and followers to study their opinion.
We collected 320,450 Tweets posted about “Odd
even” urban experiment
63,988
Data: Public Opinion
08
10. Using Google Traffic API, we track congestion
levels of main roads in Delhi.
Data: Traffic
Main Roads in Delhi Main Roads from outside Delhi 09
11. Data: Air Quality
We collect data from different sources:
• U.S. Embassy Air Quality Monitoring Station
• DPCC Program (Delhi Pollution Control
Committee)
10
13. Ground Reality
Traffic did improve
Public Transportation did handle demand
Based on our tests, and confirmed by many studies [2][3]
Based on our study, and confirmed by many studies[1]
12
[1] Business Standard: Odd-even plan: Traffic is less, pollution level isn't http://goo.gl/La4ReX
[2] Indiatimes: Odd-even plan shows no strong effect on Delhi pollution http://goo.gl/LLpkJE
[3] IndiaExpress: Week after odd-even, pollution levels spike in the capital, says CSE http://goo.gl/e5WV3b
[4] Odd-even policy: Metro ridership sees 11 per cent increase http://goo.gl/U1cnPR
[5] Odd-Even Plan: 4 Million Commuters Estimated To Have Travelled In DTC Buses http://goo.gl/sPCtoU
Air Quality did not improve
Based on many studies [4][5]
15. We label users as AAP supporters, BJP
supporters, apolitical users, and bi-political
We use different features:
• Tweets
• Mentions
• Friends
Inference of political leaning
14
16. Training
• Use users’ biographies to
manually identify those
who clearly express their
political affiliation.
Prediction
Inference of political leaning
0
5000
10000
15000
20000
25000
30000
AAP BJP Bi-political Apolitical
“I am an architect (IIT Roorkee)
by profession & a rebel and activist
at heart! and yes a pucca bjp
supporter.“ @Chhabiy
15
Numberofusers
17. Volume of tweets over time
Distribution of tweets volume over time
Before experiment After experimentDuring Experiment
The beginning of the experiment
After a long weekend, supporters
claimed the success of the event after
3 days
People’s reaction about Arvind kejriwal
speech [Watch Video].
16
18. Distribution of tweets by topic
Takeaways
• Air Quality is the most
discussed topic followed by
congestion.
• Public transportation topic was
trending at the beginning of the
experiment.
17
20. Air Quality: Positivity by camp
FirstWeek
BeforeExperiment
SecondWeek
BeforeExperiment
FirstWeek
DuringExperiment
SecondWeek
DuringExperiment
FirstWeek
AfterExperiment
SecondWeek
AfterExperiment
FirstWeek
BeforeExperiment
SecondWeek
BeforeExperiment
FirstWeek
DuringExperiment
SecondWeek
DuringExperiment
FirstWeek
AfterExperiment
SecondWeek
AfterExperiment
Beforeexperiment AfterexperimentDuring Experiment Beforeexperiment AfterexperimentDuring Experiment
Distribution of opinionated users over time Fraction of opinionated users over time
AAP
BJP
Apolitical
Bipolitical
Negative
Positive
19
21. Users
• First graph captures the volume of opinionated users as well as the negativity and positivity in every
camp.
• Second graph captures the percentage of negativity and positivity in every camp to visually identify
majority.
Takeaways:
• Most of users have negatively discussed Air Quality topic.
• Bipolitical users were more engaged in the discussion, it might be related to the fact that these people
are journalists and social activists.
• AAP supporters were more optimistic before the experiment.
• The number of AAP supporters who were engaged in the discussion start decreasing starting from 1st
week of the experiment.
20
Air Quality: Positivity by camp
22. Public Transportation: Positivity by camp
FirstWeek
BeforeExperiment
SecondWeek
BeforeExperiment
FirstWeek
DuringExperiment
SecondWeek
DuringExperiment
FirstWeek
AfterExperiment
SecondWeek
AfterExperiment
FirstWeek
BeforeExperiment
SecondWeek
BeforeExperiment
FirstWeek
DuringExperiment
SecondWeek
DuringExperiment
FirstWeek
AfterExperiment
SecondWeek
AfterExperiment
Beforeexperiment AfterexperimentDuring ExperimentBeforeexperiment AfterexperimentDuring Experiment
Distribution of opinionated users over time Fraction of opinionated users over time
AAP
BJP
Apolitical
Bipolitical
Negative
Positive
21
23. Users
22
Public Transportation: Positivity by camp
• First graph captures the volume of opinionated users as well as the negativity and positivity in every
camp.
• Second graph captures the percentage of negativity and positivity in every camp to visually identify
majority.
Takeaways:
• Discussion about public transportation start trending during the experiment, despite the fact that it
wasn’t the main focus of the experiment.
• Surprisingly, BJP supporters were more positive about the experiment during and after the experiment.
24. Traffic: Positivity by camp
FirstWeek
BeforeExperiment
SecondWeek
BeforeExperiment
FirstWeek
DuringExperiment
SecondWeek
DuringExperiment
FirstWeek
AfterExperiment
SecondWeek
AfterExperiment
FirstWeek
BeforeExperiment
SecondWeek
BeforeExperiment
FirstWeek
DuringExperiment
SecondWeek
DuringExperiment
FirstWeek
AfterExperiment
SecondWeek
AfterExperiment
Beforeexperiment AfterexperimentDuring ExperimentBeforeexperiment AfterexperimentDuring Experiment
Distribution of opinionated users over time Fraction of opinionated users over time
AAP
BJP
Apolitical
Bipolitical
Negative
Positive
23
25. Users
24
Traffic: Positivity by camp
• First graph captures the volume of opinionated users as well as the negativity and positivity in every
camp.
• Second graph captures the percentage of negativity and positivity in every camp to visually identify
majority.
Takeaways:
• Discussion about traffic and congestion start trending during and after the experiment, as people start
commenting and sharing their experiences.
• AAP/BJP supporters were equally satisfied during the experiment, but majority of BJP supporters were
more positive evaluating the experiment.
30. Users
29
• We identified place of users manually by checking their biographies, and grouped them to 4 groups
{Delhi, India, WorldWide, unknown}. In this study, we are interested in 3 groups {Delhi, India,
WorldWide} in this study.
• The graph captures the volume of opinionated users as well as the negativity and positivity in every camp
to evaluate the engagement of users, in every week.
Takeaways:
• BJP supporters in India engaged in discussing Traffic more than those in Delhi, due to the fact that BJP
is a big national party.
• Number of BJP supporters outside India who were engaged in discussion is higher than those in India.
• BJP supporters outside India were more positive during the experiment than those in India.
Traffic: Topics, political affiliations &
locations
31. Users
• Based on our tests, we found that Traffic has improved, but Air
Quality did not improve
• Based on some studies public transportation handled demand.
• AAP have been more positive about the experiment compared to
BJJP
• People were overall satisfied with the service provided by public
transportation.
• BJP supporters are more ideological compared to AAP supporters.
Conclusion
30