Political Marketing
"Political Marketing (Shama 1975) aims to shift public opinion in a matter of debate towards a desired direction. Such promoted debates are highly popular on social networks and can be massively founded.
In some cases, these efforts even involve organized and automated systems. Given the amount of money invested and hundreds of millions of users as possible targets, it is necessary to improve the understanding on the propagation of political messages in social networks."
"Using sentiment analysis to predict opinion inversion in Tweets of political communication."
Matalon, Magdaci, Yamin ; Scientific reports 11.1 (2021): 1-9.
Unprecedented Implications On Everyone Life
Elections Campaigns
Activity on Twitter has a positive influence on political candidates [3], making users more likely to vote [4] and can even be used to predict election results [5]. Naturally, these led to huge investments in PMSN, such as the $900M invested during the 2018 US midterm elections [2].
Political Debates
This includes event-based activity, such as the Brexit [6, 7], Covid vaccines, or the Eurovision contest [8]; as well as ongoing conflicts,
climate change, immigration, and the case-study of this research - the Israeli-Palestinian conflict.
Modern Warfare
Either as a recruitment platform as used during the Arab spring [9]; or as a digital battlefield, such as the Russian intervention in the US elections. These cases often involve organized and automated systems and fueled with fake news [10, 11, 12].
Political Marketing Content Propogates Differently
PMSN has three characteristics that differentiate it from the traditional marketing of products and services. Out work is the first to consider these elements and the first to ever model Debate Dynamics, with the ability of users to invert others' content.
The Users
Everyone has their own opinion. Influence lies among the neutrals.
Network Topology
It is an echoed world. People tend to connect with those who share their opinion.
Debate Dynamics
Each message can be inverted, inciting new threads among the competition.
There are hundreds of millions of users Some users are biased, others may be already persuaded
We aim to reach as many neutral individuals as possible
​
Whom should we target?
Political Marketing as a Transmission Process
Transmission models have been used to describe the propagation of a wide variety of contagion processes, including infectious diseases, viral marketing, information spread, and more. These models can be useful for evaluating the influence of a node within a social network.
An individual is succeptible to become infected
An individual got infected and can infect others
Recovered if immune from the disease (or removed from the population)
All Twitter users are assumed to be susceptible as they can all be reached through the network
A user shared a tweet seen in the feed. His followers may share it now as well.
The tweet is practically disappeared from the feed and not available for followers.
A Complete Social Networks Mining Framework
Three Hops Network
Filtering Irrelevant Tweets
Using a dedicated Machine Learning algorithm [ref] for filtering irrelevant tweets.
Did you know that 'SJP' stands both for 'Sarah Jessica Parker' and an anti-Israel movement?
Overall, 715K relevant tweets remained.
Tweets extraction
We used Twitter API to extract millions of English tweets which contained keywords to be relevant to our domain
Network Development
We first identified a set of 208 domain activists, who frequently publish content about the Israeli-Palestinian conflict. They are the seed users of the network.
Then, based on their social relations, we extracted the first and second hops of followers.
Overall, the network contains 159K users.
Understading Users' Opinions
By two ML models. The first combines Nature-Language-Processing (NLP) modeling with network-based features and is able to classify user opinion with 92% accuracy.
The other relies solely on network features, achieving 80% accuracy overall.
Identifying Influencial Users
Explore content propagation to identify the optimal seed users that maximize exposure among targeted users.
Transmission Model Simulations
Each user functions as a source. Using our network and ML models, we are able to execute extensive simulations on real-life settings, evaluating each user's diffusion potential.
Our KPI for diffusion is the number of neutral users exposed to the thread, minus the number of users exposed to inverted threads.
Introducing Content Inversion
It's A Whole New Game
Just as true for pathogens, tweets have shown to evolve over time [1]. Their intent may be mutated during an adoption, in the form of Quote tweets.
User behavior on political debates on Twitter seems to be strategic:
​
When users' opinion regarding the conflict is aligned, they can share messages and amplify their spread through the network. However, when the content is being watched by opposing users, that is, from the other side, they may use it to inflate a rival thread supporting their side of the debate.
There will never be an "Autonomous Zone" in Washington D.C..."
123.7K
30.3K
We've placed a public interest notice on this Tweet for violating...
How does Content Inversion work?
01
A user posts a tweet.
02
His supporting followers can like it, retweet it as-is, or share it with a comment - Quote
03
Opposing users invert the tweet - share it with a comment that matches their agenda
04
Each inversion triggers a new thread, advocating the other side of the debate
​
When not considered properly, the inversion can cause counterproductive outcomes
Without invertion
Total exposed: 26,244
Total adopters: 1,206
Total inversions: 0
NTE: 22,177
With Invertion
Total exposed: 70,692
Total adopters: 1,090
Total inversions: 2,307
NTE: -17,457
It's an Ech ed World
Tell Me What Your Friends Think And I'll Know What YOU Think
The opinions of connected are highly correlated (~0.6)
This forms two clusters with high intra-cluster connectivity
& low inter-cluster connectivity, that is, Echo Chambers[1].
The neutral population acts as a bridge.
​
Log of occurrences of links within the network
​
​
What happens Within a Chamber Usually Stays in the Chamber
As closer we are to the core of this debate, we identified very dense communities with a 1000X higher cluster coefficient than the average on Twitter. Segregation between the sides is clear.
​
Visualization of the developed network made on Gephi. Israeli supporters are presented in blue, opponents are presented in red, and neutrals are presented in grey. Darker tunes for each side correspond to stronger levels of polarity.
Diffusion as a double-edged sword
The synergistic interaction between the network topology and the debate dynamics empowered the effect of echo chambers in our network.
These chambers function as a supportive environment that amplify the source. In addition, allows the content to propagate to new areas in the network, unreachable without supporters' adoptions.
On the other hand, content can be inverted by opposing users, resulting in the other side overtaking the diffusion.
On LOW transmission settings
Campaigners should target users with many neutral followers, such as super-spreaders. Adoptions may be beneficial, but their gain is limited, mainly because of the echo chambers and the low transmission.
On HIGH transmission settings
On high transmission settings, a trade-off is presented; on the one hand, one should have a dense and supportive environment that serves as an amplifier. On the other hand, such propagation may serve as a cause of inversions.
References
-
Matalon, Yogev, et al. "Using sentiment analysis to predict opinion inversion in Tweets of political communication." Scientific Reports 11.1 (2021): 1-9.
-
Edelson, Laura, et al. "An analysis of united states online political advertising transparency." arXiv preprint arXiv:1902.04385 (2019).
-
Kruikemeier, Sanne. "How political candidates use Twitter and the impact on votes." Computers in human behavior 34 (2014): 131-139.
-
Miller, Carl. "The rise of digital politics." (2016).
-
Terán, Luis, and Kidus Yirgu. "Estimating Public Opinions Using Twitter Data: The Case of the 2018 Ecuadorian National Referendum and Constitutional Reforms." 2019 Sixth International Conference on eDemocracy & eGovernment (ICEDEG). IEEE, 2019.
-
Bessi, Alessandro, and Emilio Ferrara. "Social bots distort the 2016 US Presidential election online discussion." First monday 21.11-7 (2016).
-
Gorodnichenko, Yuriy, Tho Pham, and Oleksandr Talavera. "Social media, sentiment and public opinions: Evidence from# Brexit and# USElection." European Economic Review (2021): 103772.
-
Liu, Zhe, and Ingmar Weber. "Is Twitter a public sphere for online conflicts? A cross-ideological and cross-hierarchical look." International Conference on Social Informatics. Springer, Cham, 2014.
-
Nasrallah, Antoine, and Nada Sarkis. "The Role of Social Media During the Arab Spring." Business and Social Media in the Middle East. Palgrave Macmillan, Cham, 2020. 121-136.
-
Bovet, Alexandre, and Hernán A. Makse. "Influence of fake news in Twitter during the 2016 US presidential election." Nature communications 10.1 (2019): 1-14.
-
Rofrio, Daniel, et al. "Presidential elections in Ecuador: Bot presence in twitter." 2019 Sixth International Conference on eDemocracy & eGovernment (ICEDEG). IEEE, 2019
-
Obadimu, Adewale, et al. "A comparative analysis of Facebook and Twitter bots." Proceedings of the Southern Association for Information Systems Conference, St. Simon‟ s, Island, GA, USA. 2019.