What India’s SOS tweets reveal about its deadly second wave

Citizen-driven organisation and mutual aid emerged on Twitter after a failed response from the Indian government to a public health emergency.

AP

As India reports an eye popping number of cases every day during its worst phase of the pandemic, hospitals have been forced to turn away patients as critical supplies of oxygen and medicines run short.

Media outlets in the country and around the world drew attention to the horrific condition of the country’s public and private health infrastructure, which has been pushed beyond the brink.

In response to the unprecedented public health emergency, people have taken to bypassing conventional lines of communication and turned to social platforms like Twitter to crowdsource help for hospital beds and oxygen cylinders.

Research conducted by the Atlantic Council’s Digital Forensic Research Lab (DFRLab) explored how Indians used Twitter to coordinate resources for one another in the absence of a robust state response to the health crisis.

Using Twint, an open-source intelligence tool, DFRLab analysed the activity of SOS tweets by Indian users between March 1 and April 21.

A combination of more than 20 keywords related to the most common medical resources (oxygen, plasma donation, hospital beds) were paired with combinations of words associated with assistance (help, require, needed) to filter out relevant SOS tweets.

In total, 81.63 million tweets were yielded from users directly asking for help or responding with a relevant link over the timeframe.

The most requested medical supplies were related to sourcing or refilling oxygen cylinders, accounting for 36.9 million of the total tweets. Antiviral drug Remdesivir was the next most requested with 14.1 million tweets, followed by 13.9 million tweets for procuring hospital beds.

Other

While a total of six million SOS messages were recorded during the month of March, the dataset found the number of SOS tweets increased over sevenfold to 41 million between April 8-14.

On April 8, the country had recorded its then-highest single day increase of new infections with 131,787 cases. Over the coming weeks, that number would skyrocket to over 380,000 new daily cases.

DFRLab parsed the dataset to calculate the number of unique accounts driving the conversations around the SOS tweets, and found that between March 1 and April 21, nearly 520,000 individual accounts actively engaged with emergency tweets from other Indian users to help provide relevant Covid-related information or medical aid.

Over 356,000 of those accounts belonged to ordinary citizens, far outweighing that of prominent social media influencers or public figures.

Other

“This highlights how hundreds of thousands of regular Indian Twitter users rose to the fore in response to their compatriots’ please for assistance on the platform, and mobilized into decentralized support networks to help source information and medical supplies through conversations on Twitter,” wrote the authors of the DFRLab analysis.

Most notably, this decentralised web of citizen-support networks continues to engage on the platform despite having multiple SOS tweets taken down by Twitter after the Indian government, led by Prime Minister Narendra Modi, ordered the platform to remove posts critical of its handling of the virus.

Many on social media criticised the government for focusing on “censorship” while the country was in the midst of a “humanitarian disaster”.

The data also showed the geospatial distribution of the tweets across the country. The Delhi National Capital region witnessed the highest number of requests for assistance, with 39.86 million SOS tweets.

Other

DFRLab also made a clarification regarding the digital divide reflected in the data. While Twitter is popular with an estimated 17.5 million users in the country, it only represents 1 percent of the Indian population.

As a result, the hotspots analysed through the SOS tweets is only one half of a much larger picture, given cities with poorer internet connectivity and higher caseloads are absent from the dataset.

Route 6