BJP’s Twitter network decentralised, Cong’s unstructured
Researcher worked with 1.7 billion tweets and wrote algorithm to filter through spam by spotting source of trend
An independent researscher recently conducted an intensive study on Twitter posts over the past couple of months in an attempt to understand how the BJP and Congress, the two largest political parties in the country, spread information, or misinformation, on Twitter.
The most interesting inference is that the Twitter networks of both parties is very dissimilar: The BJP’s network is very decentralised while the Congress’ is unstructured. However, the scale of both networks is quite large, and comparable. A few weeks ago, the researcher, who asked to be identified only as Kumar, had conducted a study on the most active Twitter accounts that posted political content. He found that 85 per cent of followers of the most popular accounts were bots — pre-programmed accounts which work without manual intervention. His latest study is a continuation of the earlier one.
Kumar worked with a database of over 1.7 billion tweets, which contained spam propaganda content pertaining to all political parties in the country, made by bot accounts since the beginning of November. Through an automated algorithm, he had identified them by parsing through the most popular political accounts on Twitter.
For this study, Kumar wrote an algorithm to filter through the spam by identifying the source of a particular trend. He called these accounts “seeds”. Another category of accounts he identified were those which tweeted at abnormally high intervals; for instance, accounts which tweeted over eight lakh times in seven years.
Speaking to Deccan Chronicle, Kumar said he wished to understand how propaganda and misinformation is spread on the internet.
“My study identified political content by most parties in the country. However, in terms of scale, the Congress and the BJP have the biggest. Both parties’ Twitter networks work very differently,” he said.
His algorithm identified over 2.7 lakh accounts which tweeted BJP content. As many as 7 per cent accounts were seeds. He also prepared a network map using this data. “The BJP’s network maps shows a very methodical approach. It has individual clusters which are almost islands of misinformation. Each cluster has a seed which tweets the source material while the other accounts retweet it. There is very little communication between the clusters,” Kumar said.
On the other hand, a lack of structure in Congress’s network of over 1.2 lakh accounts was evident from its network map. Of these accounts, only 0.1 per cent were seeds. The map shows a blob of activity, with seed and other accounts jumbled in between.
“The Congress, it seems, is still at the stage of taking baby steps. It does not have the BJP’s structure,” Kumar said.
Another way both networks differ greatly is the manner in which seed accounts are dealt with. In case of the BJP, when a seed account gets deactivated, another account in the same cluster is “promoted” to seed status. “In the Congress’ network, when a seed account gets deactivated, the accounts dependent on it become inactive,” he added.
One of the most important findings of the study was that the accounts of Congress office-bearers such as Rahul Gandhi never followed the party’s seed accounts. On the other hand, all seed accounts of the BJP were followed by the BJP office-bearers, including ministers.
Also, the Congress’ bot accounts did not tweet abusive content, unlike those of the BJP’s. Kumar said that the BJP was normalising hate and abuse through its leaders following such accounts.