The entire programme for the African Investigative Journalism Conference at the University of the Witwatersrand looks interesting, but I would make a bee-line for the panel discussion on “#GuptaLeaks: chasing down the fake sites and bots”, scheduled at noon.

Here, three renowned fake-account busters, Jean le Roux, Kyle Findlay and Andrew Fraser, will participate in a panel discussion to discuss their detailed research into social media manipulation.

It is a subject in which I have an intense interest. The purpose of this column is to show why everyone should – even though they may never have opened a social media account and may still rely on print and radio for their daily news.

“Fake sites and bots” are a topic of crucial importance to everyone interested in defending the free exchange of ideas on which democracy relies.

The extent to which social media (especially fake accounts that manipulate facts, express rabid opinions, and manufacture bogus outrage-communities) have reshaped the public narrative, and polarised societies worldwide since 2016, is nothing short of terrifying.

I have done a lot of reading on this topic and can particularly recommend an analysis by the Computational Propaganda Research Project at Oxford University.

The large-scale study, conducted in 48 countries, concludes:

“The manipulation of public opinion over social media platforms has emerged as a critical threat to public life…

“While the internet has certainly opened new avenues for civic participation in political processes – inspiring hopes of a democratic reinvigoration – the parallel rise of big data analytics, “black-box” algorithms, and computational propaganda, are raising significant concerns for policymakers worldwide. In many countries around the world, divisive social media campaigns have heightened ethnic tensions, revived nationalistic movements, intensified political conflict, and even resulted in political crises – while simultaneously weakening public trust in journalism, democratic institutions and electoral outcomes.”

Today, Le Roux, Findlay and Fraser will be analysing this phenomenon in the South African context. Their research holds particular interest for people like me, who naively embraced social media (especially Twitter) as a magnificent example of the way technology could extend democratic participation, by linking individuals and interest-based communities in “good faith” group discussions online.

Their conclusions also add a new dimension to the crisis facing the mainstream media (known as MSM in social media shorthand).

There has (rightly) been much rending of garments and gnashing of teeth over the extent to which some gullible newspapers allowed themselves to be used by “whistle-blowers” to plant false scandals to destroy Zuma’s “enemies” and “capture the state”.

Those responsible for these “leaks” were real human beings driven by personal agendas (and maybe additional incentives) to steer the public debate in favour of the ”Zupta” gang and its State Capture project.

We will never know the full story, because the newspapers involved refuse to reveal their sources, even though they planted falsehoods. But that was last week’s story.

This week the focus is on how old-fashioned manipulation tactics pale into insignificance compared to the threat posed by big data analytics, black box algorithms, and computational propaganda.

Not even those who steer clear of social media are immune.

In a frantic bid to remain relevant in a changing communications environment, print and broadcast media (MSM) have followed social media trends like a pack of dogs pursuing a bitch on heat.

What’s more, the new generation of journalists rarely read newspapers themselves, relying primarily on social media to define the issues they report on in the MSM.

Many of them are also recent graduates, fresh from tertiary institutions where they were fed a staple diet of “grievance-and-victimhood studies”, underpinned by “identity politics” that seeks to divide and polarise society on the basis of race, sex, sexuality and other immutable biological factors.

In this ideologically driven universe, the greater your victimhood, the higher your social standing.

This is fertile terrain for big data analytics and algorithms to create a range of online “identity tribes”, competing with each other in a so-called “victimhood Olympics” and in hunting down so-called “oppressors”.

Young people (including many communications “professionals”) seeking social affirmation and acceptance demonstrate their “tribal loyalty” online by denigrating their opponents and turning the pernicious “identity” ideology (on which apartheid was based) into “progressive discourse”.

A great deal of the most vitriolic online content is maliciously created by sockpuppets (real people who mask their identity) who then link their posts to a network of “bots” (fake accounts) that amplify the original post and inject it, like a virus, into broader conversations targeting real people. Before long, the bots (even where they are a minority in a conversation) overwhelm these interactions by mocking or marginalising moderate views.

The resultant “controversy” (depending on its magnitude) is then faithfully reported in the MSM, desperate not to remain the only virgin in the orgy.

This phenomenon first surfaced in a Sowetan article based on research by Jean le Roux (@jean_leroux) on the influence of the Zupta Twitter network.

It was then analysed in detail by bot-sleuth Andrew Fraser (@Arfness) when, in 2016, he exposed the network of some 800 fake accounts, many emanating from “troll farms” in India, called #Guptabots.

They sought to blame “white monopoly capital” for South Africa’s problems and promote “radical economic transformation” as the solution.

In 2017 this phenomenon was linked to Bell Pottinger, the disgraced London PR firm hired by the Guptas to support their bid to ensure Jacob Zuma was succeeded by his ex-wife, in order to defend their “state capture” project.

This revelation caused a furore, but was merely the tip of the iceberg.

Although by then many of us realised that something was afoot, it was frustratingly difficult to pinpoint and quantify – until Twitter and Facebook (under intense public pressure over the extent of the manipulation of their platforms in the run-up to the 2016 Presidential election in the United States) – were forced to begin cleaning up their act in order to save their businesses.

This led Twitter to undertake a massive bot-cull (suspension of fake accounts). It is important to note that this did not include all fake accounts. There are so many millions of them that removing them all would have destroyed Twitter’s business model, which depends on its capacity to effect widespread influence. (The fact that accounts are fake does not detract from their potential influence. In fact, their poison often enhances their potency.)

So Twitter has only removed what it described as the most vicious and egregious fake accounts. Overnight, my own followers dropped by approximately 25,000.

The tone of my timeline changed dramatically, for the better. But it is still undoubtedly infected by hundreds of thousands of fake followers, classified by Twitter in the second league of vitriol. It has proved very difficult to establish just how many there are, because identifying bot accounts is a complex undertaking.

I have tried to use various “apps” that claim to distinguish between real and fake Twitter followers. According to “TwitterAudit,” a full 616,329 of my 1.35-million followers are fake. Whether or not this is true, there are definitely a lot of “bots” trying to negatively influence conversations and opinions involving me.

For the purposes of scientific research, data scientist and bot-buster, Findlay (@SuperlinearZA), could only use officially identified bots, suspended by Twitter. This sample enabled him to identify the prevalence of these fake influencers in major online controversies over the course of the past four years. By using a number of “control factors” Findlay concentrated primarily on three issues in which suspended accounts were particularly active:

• the run-up to ANC’s 54^th elective conference, where a veritable bot-army sought to boost the candidacy of Nkosazana Dlamini Zuma;
• the response to my “legacy of colonialism” tweet; and
• the #BlackMonday anti-farm killings march.

Having read Findlay’s path-breaking paper, I now understand how the bot community injects itself into conversations, and creates new ones, using my own case as an example. They target me in their posts, hoping I will respond. I often do, naively believing that I am reacting to legitimate queries or criticism.

These “needling” tweets are merely the “bait” often used in more conventional e-mail phishing expeditions. When I respond to a bot account, I unwittingly give it credibility among my hundreds of thousands of genuine followers, many of whom then also engage, creating a new “conversation community”, and opening the door for hundreds of new bots to overwhelm the conversation to vilify my supporters’ views. Most genuine people, with real feelings, quickly back out of such encounters. So the algorithms win.

As in all comparable cases, the purpose is to disparage me, and through the pervasiveness of social media, ensure that the “bot-line” becomes the dominant narrative, supported by its further dissemination through “captured” newspapers, such as the ironically named “Independent” group.

This has helped me to understand what happened on 16 and 17 March 2017. According to Findlay’s analysis, over those two days alone, the proportion of hostile tweets aimed at me that were linked to suspended users (bot accounts) was higher than the proportion of tweets by the “bot army” that sought to influence the outcome of the ANC’s 54^th elective conference (over a much longer period of time).

And that wasn’t even counting the involvement of the tens of thousands of accounts on my timeline that are probably also fake, but have not (yet) been suspended by Twitter!

This is not to suggest that there wasn’t a degree of genuine outrage and disagreement with my series of tweets on the lessons I learnt during my visit to Singapore. (There is always some genuine outrage underlying a fake community.)

But there is also no doubt that this reaction was massively exaggerated, through thousands of bots that decontextualised and distorted my comments with the purpose of demonising me and dividing the DA.

This all occurred before the Bell Pottinger revelations, or the Mueller Indictments (in the United States), so most of us were unaware of the extent of social media manipulation behind this story. The MSM, including radio and television, reported it as if the controversy was genuinely rooted in real community outrage on a massive scale. The result, as Findlay says, was “unprecedented”.

His research also throws up sub-themes that are equally fascinating, such as the “conversation communities” created by real people, such as Mbali Ntuli and Eusebius McKaiser (the world debating champion and broadcast-equivalent of Iqbal Survé), to fan the flames.

But I will leave that analysis for another day.

As the research has shown, the political purpose behind these tactics is not merely to create “villains” in a narrative. It is to divide and paralyse organisations.

The extent to which this often succeeds is frightening and raises a number of questions, not only for the media, but also for political parties.

In my own case, the obvious question is: why would the real people behind sockpuppets and bots even bother? I resigned as DA leader in 2015, and am on the final lap of my last term as Premier.

Why would social media, that primarily targets generation of tomorrow, focus on “yesterday’s woman”?

I now understand that the purpose is to sustain a narrative about the DA.

Because our opponents have no other arguments, they default to the race card. Despite overwhelming evidence to the contrary, they dismiss us as a “white party” defending “white interests”.

This line of argument was dealt a serious blow when I was succeeded by a black leader. So, from the get-go, our opponents sought to propagate the vicious (and racist) lie that he was my “puppet”.

While I was trying to phase myself out of the political arena, the DA’s opponents were desperate to keep me alive, and portray me as the evil white controller, hovering on a broomstick over every black leader and office bearer in the party.

Social media became their weapon of choice – made worse by many in the MSM who regard the DA as a legitimate target. (Some journalists still have an interest in exposing the media manipulation targeting the “good guys” in the ANC, but that courtesy tends not to extend to the DA.)

Before long the propaganda started reflecting in polling results and focus-group discussions, which goes a long way to explaining the DA’s response to me.

We all know that, in this kind of environment, electoral outcomes will inevitably be disproportionately shaped by those who run the best online disinformation campaigns. And in a fragile society, battling to heal race divisions, it is no surprise that the manipulators focus primarily on deepening racial divisions for the purpose of electoral gain.

Those of us trying to communicate an alternative vision of an inclusive South Africa, fairly and honestly, will be running the electoral ultra-marathon in leg irons.

To think that a political party, on its own, can neutralise this manipulation machine is like expecting a single teetotaller to challenge the market influence of South African Breweries.

The only place to stop it is at source.

We need a massive, global campaign to pressure Twitter to cull every fake account, not only the most vicious, in order to limit the mass algorithmic manipulation of information and opinion.

The stakes could not be higher. If we don’t, it is no exaggeration to say that the social compact underlying our Constitution (and that of many other democracies) could unravel. DM