Who’s tweeting whom?


The rise of Twitter as a  mass communication medium has confounded analysts and users alike. The then limiting 140 character-size seemed almost counter intuitive. Despite this Twitter has gained traction to become the third most widely used social media microblogging platform in the world. It is small wonder that Twitter has become a target and a medium for propaganda and cause-related matters.

McLuhan in 1970 a decade before the proliferation of computers ominously predicted the next World War will be a guerrilla information war with no discernible division between military and civilian participation. This turned prophetic as the global digital village has elevated every online poster to a potential influencer. Such information has become an integral component of the arsenal in warfare.

We were intrigued of about Twitter in the South African context and undertook a long term study of the #FeesMustFall Twitter campaign, looking at about 600 000 tweets as part of my Masters in IT under the guidance of supervisor and co-author Colin Thakur. We  specifically examined the sentiment of tweets with a view of determining if there was a correlation between real-life events and Twitter sentiments. More on this Sentiment Analysis in another contribution.

While examining tweets we discovered some tweet activity which could not have been undertaken by a  human. This then led to our discovery of software robots commonly called bots deployed in the #FeesMustFall campaign right from the beginning of the campaign in October 2015.

A software robot  or bot is an automatic or semi-automatic computer algorithm that automatically produces content, mimics humans or interacts with humans on social media. A bot is amoral. It exhibits the intent of the developer who themselves may be agenda-based or disinterested hired guns. A cyborg is a semi-automotive bot that is either bot-assisted human or human-assisted bot.

Marketers, lobbyists, and activists use or deploy bots to amplify their agenda by attempting to manipulate social media users’ opinions through automated social engagements. The agenda may be personal, cause-related, political or commercial. They are not always bad as they are often benign or useful, although some are created to harm, by tampering with manipulating or influencing and deceiving social media users. 

One study found that as many as 400 000 social bots participated in the 2016 US elections, supporting both President Trump and contestant Clinton. These bots generated at least 3.8m tweets or 19% of the total tweets during the observation period. They also showed how political bots were used in the US context to surreptitiously coordinate campaign, and even illegally solicit contributions or votes. The goal in these examples are overt and covert social media manipulation.

Is Twitter important? Twitter has been used in many-to-many crises communications where the minimalist character-limit mitigated emotion while simultaneously encouraging factual information. This microblogging feature was leveraged by the Japanese in 2011 when they endured a combo of an earthquake, followed by a tsunami, all but collapsing the communication infrastructure except for Twitter and Facebook. The Queensland Floods (2010) and Christchurch earthquakes (2011) further demonstrated the Twitter platform usefulness which assisted search and rescue. The now customary location feature further supports first responders.

The sentiments expressed on Twitter may well serve as a barometer of the mood of the nation. However the existence of bots “clouds” this assertion.  The fact remains that Twitter is now certainly as important as the 4th estate (some call Twitter the 5th estate) given that daily newspaper sales account for  just over 1.05m newspapers per day which equates to about 5m readers  while  Twitter has 8.3m South African users justify our research platform choice.

The GuptaLeaks controversy saw programmers creating an army of agenda-driven bots which targeted folk critical of the Gupta family. This is now called the weaponization of social media. This means that sentiment analysis research should and must take cognisance of the fact that the tweets often though not always have an element of rogue contributions.  Further social media opinion may be influenced by what is now called ‘computational propaganda’.

An academic, rightly questioned the need for bot research in the #FeesMustFall campaign, because they already knew bots played a role in the #FeesMustFall campaign. I listened in dismay until Dr Thakur listened and exclaimed “Science is about proving the obvious. The researcher who scientifically proves a new fact considers this new knowledge to be profound. If presented articulately, the first person who hears this new fact considers this revelation logical. The rest of humanity should consider this revelation ‘common sense’. 

This is your goal – to make every reader think that you profound discovery is common sense! Think, cloud, think, rain!” This provided me with much comfort in the journey. I also could find no evidence of bots in the campaign in popular media or academic journals. With Twitter shutting down up to 70 million fake or suspicious accounts in July 2018, this work on the role of bots has, grown in relevance.

So how does one identify a bot? The following are some typical bot characteristics of Bots: 

1.    Follow very few accounts and in turn are followed by very few tweeters.

2.    Are usually topic or cause-specific.

3.    Exhibit regular repetitive behaviour and tweet the same thing all the same time.

4.    Display excessive duplicate or near-duplicate tweets on profile’s stream.

5.    Tweet Frequently

These are the kinds of characteristics one should look for in deciding if a tweeter is a human a bot or a cyborg. A real quick test is to use Botometer created by Indiana University. If this research intrigues or overwhelms you, do contact the authors. A research we will next undertake involves algorithmic techniques and tools to detect and limit the mass algorithmic manipulation of information and opinion.


This piece was written by Yaseen Khan and Colin Thakur who both acknowledge the active support and guidance of Minster Stella Tembisa Ndabeni-Abrahams, who was fascinated this research as it fell within her goal of #BuildingACapable4IRArmy