The Indraprastha Institute of Information Technology (IIIT) Delhi's Laboratory for Computational has launched a platform to track hate speeches on the microblogging site Twitter during Assembly Elections 2022.
The platform 'Project Robinhood' includes student members, namely Drishya Uniyal, a master's student, Sarah Masud, a PhD student, and Dr Tanmoy Chakraborty, assistant professor. As geotagged by the tweets, the team has so far found that Uttar Pradesh, Delhi, and Uttarakhand reported the highest hateful content based on the data curated.
The students said that most politicians, leaders, and affiliates use social media platforms to share updates of their political campaigns during elections and make announcements about contesting candidates, besides criticising opposition party members.
"To analyse these linguistic patterns from the point of hateful content, members of LCS2 at IIIT-Delhi launched this project," students told News18.
Tracks Content On Twitter
The project will track content on the Twitter platform for the Assembly Elections 2022 and curate information from January to March. To share the insights with a broader audience, the team has launched a portal that provides aggregated and granular statistics of the various groups of users generating political hate and the groups it is targeted against. This information is further analysed at the state level.
The data collection for the project started in January this year, beginning with 50 political leaders from five significant parties — BJP, INC, BSP, SP, and AAP. In addition, the official state-wise Twitter handles of these parties are also being tracked. Based on the curated data, the team found that Delhi and Uttarakhand reported the second-highest hateful content after Uttar Pradesh.
"This is an interesting observation as Delhi is the nodal centre for three of the country's major parties. Despite the national capital not conducting elections, it still produces a significant volume of politically hateful tweets," the students added.
Rise In Hateful Content
Another observation the team made is that direct attacks in political parties prevail over attacks on individual politicians. The students attributed this to calling out the party's name and creating more engagement from the opposition party and ordinary citizens. Around January end, there was a rise in the hateful content the team analysed, which was due to the political rally held in Punjab during the same time.
Further, the team expects that as the frequency of physically coordinated rallies increases, the influx of political hate on Twitter will rise. The students said that the present data curation for February is likely to reveal those insights. They added that the portal's aggregated and daily statistics are updated on a regular basis.
"Based on the data curated so far, we can say that no political leader or organisation is blame-free when it comes to engaging in hateful speeches. While some leaders are more straightforward in their expressions, others prefer taunting. Despite the best efforts from the Election Commission and strict monitoring by social media platforms, these expressions of political hate continue to evade the online platforms," the students pointed out.
A small percentage of ordinary citizens start and participate in most of the trending political hashtags on Twitter. While trending, these opinions do not reflect the democratic opinion. They added that the voters need to be educated and informed to read these trending hashtags.
"We strongly feel that both common citizens and politicians need to be sensitised concerning the quality of political discourse we want to enable on the social platforms and be provided training in ethically dealing with political speech," the students said.
To analyse the different categories of political attack and hatred, the team had to iteratively analyse and formalise the content as directly or indirectly hateful. Direct attacks specifically mentioned a politician or a political party. At the same time, indirect attacks in the forms of insults have no person or party mentioned but contain contextual evidence of political hatred.
"In addition, some of the hateful content was not political but religious in nature. Given the country's long history of politics and religion, the team added religious attacks as an additional category of harm," the team concluded.