In light of recent events in India, our paper "Narrative Building in Propaganda Networks on Indian Twitter" out today at #WebSci22 sheds some light on how hyper-partisan information is repeated, especially through a selection of influencers who consistently go viral.
We curated and studied a large amount of data (over 80 million tweets in Hindi and English tweets from over 26000 politicians and 6000 influencers), for propagandist content.
For each event, we outline how we classify tweets, detect propaganda, extract narratives (ie how the propaganda is framed), and finally estimate impact of each thread. These are detailed in the full paper.
When a major event takes place, a set of influencers typically converge on a specific narrative of the event, and aggressively retweeted by their followers, but also a pyramid structure of lower-level influencers.
The same influencers repeatedly show up across major events.
Each major event has sub-narratives, each driven by a set of influencers.
With CAA, sub narratives like nationalism, politics etc, were put out, some stuck with audiences. THESE in turn get amplified.
Here is the node structure for the PFI narrrative which emerged & went viral
Once "PFI" goes viral as anti-national, associated narratives (eg Shaheen Bagh) can also follow the same narrative, since the influencers who drive them have convergent sets of followers.
Despite precipitating events of viral social media significance being fundamentally different – migrants, COVID, Bollywood etc, the same set of accounts act as connectors to turn a narrative towards an accepted dogma, that one group is a consistent villain irrespective of event.
Our work "Insights Into Incitement: A Computational Perspective on Dangerous Speech on Twitter in India" just won an honorable mention @acmcompass.
It shows why "dangerous speech", rather than "hate speech" helps understand online incitement in India, using large-scale data
1/n
Dangerous speech is actively propagated, rewarded, and endorsed in increasingly polarized, and subsequently radicalized echo chambers.
It is distinct from explicit hate speech, and machine language classifiers often miss, or misclassify it.
Our data on key polarizing events shows that "dangerous users" are more active on Twitter as compared to other users as well more influential in their networks, in terms of a larger following.
They also interact significantly with verified accounts.
The consistent anti-Bollywood trending emerges from a group of dedicated Sushant Rajput fans whose networks, history & output give insight into the radicalization of online populations & organized trolling in India.
It's also a failure story of Twitter as a platform
Thread:
The community treats the death of Sushant in the same way that religious communities treat dogma. There can be no evidential argument on it. In this it is closer to the communities that spread theories about the death of Diana Spencer than typical online fan-clubs
There are various people engaged in SSR tweeting. Some were fans before his passing, some came on after his death mid-pandemic, some simply distrust the government on other fronts and find a community with a common goal.
Each plays different roles in the conspiracy ecology
Rejoicing an Indian film winning IMDB's GOAT rating?
As someone who studies these things, we only won the population war.
You can go watch Kashmir Files to confirm if it is the greatest film ever, or I can explain with data, if that does anything for you, why @IMDb is broken
Kashmir Files has a near total top end rating. Films don't get rated as magnificent universally. Usually there is a spread of people who love, hate, and most importantly, have moderate opinions.
See for instance comparison with Godfather and Dark Night, high rated films on IMDB
This places it well above every Indian film ever made. This also puts Vivek Agnihotri in the category of Coppola, Spielberg, Tarantino, Scorcese, Kurosawa among others whose films hit that note.
We studied military veterans party engagement on Indian Twitter
TL;DR
BJP has huge advantage w/ military veterans
A small number of very political veterans are hugely influential
Veterans are politically useful, but politicians rarely tweet back when veterans ask them for help
Despite iconic Subhas Bose, military & political leadership traditionally kept apart.
Veterans eschewed electoral politics, but with regular exceptions incl Jaswant Singh, Rajesh Pilot, Amarinder, JS Aurora etc.
Since 2014, sizable increase in veterans joining politics.
BJP holds a massive advantage among veterans, some that joined the party, have significant social media footprint and influence including @Gen_VKSingh@majorpoonia@Ra_THORe etc. Military joining the party has key symbolic extension of patriotism.
UP election hashtags study: 1. Smaller alliance parties (RLD, ApnaDal etc) hashtag about larger partners (BJP, SP), don't get reciprocated! 2. BJP hashtags more positive-themed, INC most antagonistic 3. Alliance parties prefer talking about Modi, State BJP wants to focus on Yogi
We find a high degree of organization among all parties, suggesting active IT cells generating hashtags.
All non-alliance partners attack the BJP.
SP and INC avoid attacking each other, but BSP, AIMIM do attack other non-BJP parties, suggesting potential of eating into votes.
INC has focused its hashtagging mainly on women's issues coalescing around #लड़की_हूँ_लड़_सकती_हूँ
Majority of UP INC politicians (over 80%) who have used this hashtag are male, a rare case of male politicians coalescing around a strategy framed in a female voice.