Going viral: fame or doom

Katsushika Hokusai, The Great Wave of Kanagawa (1831), woodcut print, in "Thirty-six Views of Mount Fuji", Private collection. WikiArt.

What colour was that dress? Did you accept, and spread, the Ice Bucket Challenge? How many Gangnam style video clips have you watched? Will this article be read by a handful, or hundreds of thousands?

We have all, at some time or another, been part of an online viral event. Apparently out of the blue, some small often trivial tweet, video clip, or story gets a life of its own and swamps social media.

In a matter of hours or days, someone or something enjoys meteoric success and fame, or abysmal failure and doom. The latter can drive stable people to the depths of despair, even suicide. On 20 Dec 2013, as Justine Sacco was leaving London for Cape Town, she sent an infamous tweet which went viral in the 12 hours before her arrival in South Africa, blowing her life apart. Others, sometimes only on the periphery of a viral event, are driven into hiding, sleeping on a friend’s couch for many months to escape a furore.

There is a growing literature on the subject, but little coherence or agreement, and remarkable lack of insight into the phenomenon. In this and the following articles I will consider what we do know, and what we need to find out.

The phenomena

Some try to be strict in their definition of virality. For example, Nahon & Hemsley define it as “a social information flow process where many people simultaneously forward a specific information item, over a short period of time, within their social networks, and where the message spreads beyond their own [social] networks to different, often distant networks, resulting in a sharp acceleration in the number of people who are exposed to the message.”

They clarify “sharp acceleration” as meaning “that the rate at which a viral item is viewed in a specific time period is significantly larger than the previous time period.” In placing this emphasis on speed of spread, they try to exclude related phenomena such as word of mouth (WOM), memes, and information cascades, although I do not see a good rationale for doing so.

So in simple terms there is a range. Using an audience model, non-phenomenal events are like one person getting up from the audience to go to the toilet/restroom. When there is a bad speaker or act, this might increase so that there are sporadic departures. The viral phenomenon first appears when the speaker or act is so dreadful that it starts to fall apart: the number leaving begins to rise rapidly. However the intensely viral phenomenon occurs when someone shouts “Fire!”, causing mass panic for the exits.

The problem with such intensely viral events – those falling squarely in Nahon & Hemsley’s definition – is that there is no perceptible trigger (in the model, the call of “Fire!”), and no one knows how to predict the sudden sharp rise in transmission (onset of mass panic). We can therefore only know these events once they have occurred.

Are they novel?

Although the celerity and magnitude of online viral events is unprecedented, this type of phenomenon is almost certainly as old as the first cities in the Middle East. It is very hard to gauge just how viral historic events might have been without measurements, so it is not until we have more detailed economic and business information that we can observe the likes of economic bubbles.

One of the first attempts to lay bare several bubbles, rumours, and other hoaxes was the series of three volumes written by Charles Mackay (1841) who described, among others:

  • The Mississippi Bubble, in which there was wild speculation in the shares of the Mississippi Company for a couple of years before it collapsed in 1720.
  • The South Sea Bubble, a rather slower speculative scheme which built from 1710 and collapsed in 1720, but involved King George I of England, the Chancellor of the Exchequer and Bank of England.
  • Tulipmania, in which tulips became a coveted luxury item during the early seventeenth century; variegated flowers, the result of a virus infection, became even more sought-after, and prices soared. The market abruptly collapsed in early 1637, although there is controversy as to why, and how severe the impact of the collapse was.

Wikipedia lists an excellent collation of economic bubbles.

Inevitably, with very slow communications, these evolved over a timescale of months and years. The advent of the telegraph as the first means of transmitting information more quickly than a horse and rider drastically shortened the time taken for more recent events, as Standage has explored.

The Internet then provided the communications infrastructure for the first way to broadcasting near-immediate messages to many people at once, without any editorial gatekeeper, and has provided a steady succession of viral events since.

Prior to the twentieth century, popular protests and uprisings were mainly spread by word of mouth, with printing presses providing handout pamphlets to further generate concerted action. Nahon & Hemsley quote the relatively recent example of Rosa Parks, who was arrested in Montgomery, Alabama, for not giving up her seat to a white person when travelling on a segregated bus. Within three days, word of mouth, phones, and leaflets mobilised 40,000 people to boycott the bus system in protest.

One insightful experiment is described by Sampson, and was carried out by Milgram in 1968. He placed an actor in the crowds on a busy Manhattan street, and told him to stand on a corner and look up. When a single actor did this, 20% of those passing by imitated him by looking up. However when five actors did the same thing together, 80% looked up. I can remember playing similar tricks when I was at school, many years before the Internet. People have not changed that much.


Classical modelling of these phenomena is taken straight from that of epidemics of infectious disease. The basic process at the heart of this is to take an infected and infectious individual, and assume that they will transmit the disease to a certain number of others, and so the chain goes on. It is not mathematically difficult to arrive at an index of transmissibility, to allow for that to vary between individuals, build in incubation delays, and so on, to make this more sophisticated.

That said, pronouncements on the recent outbreak of Ebola are that attempts to model that, using the most sophisticated techniques available, did not accurately forecast what happened.

If we are to accept that, at some stage during a viral event, there is a sharp acceleration in the number of people exposed, that implies that the effective transmissibility changes at that moment. Whilst that can be catered for in a classical model, there is nothing to predict when that change might happen, or how the transmissibility will change.

Unravelling what actually happens during an intensely viral event is very difficult. Ni et al. attempted to do this by sampling social media for the Ice Bucket Challenge, but their own results show that it propagated far more slowly than the sort of Twitterstorm that affected Justine Sacco. So although their study is fascinating – it concludes that the Ice Bucket Challenge propagated much like the 2009 influenza epidemic – it does not help us understand more intense phenomena.

I have yet to see a published model which fits well the real data resulting from a Twitterstorm, such as those shown by Nahon & Hemsley.


There are plenty of physical events which exhibit sudden dramatic change, and you don’t have to be a surfer to think of one of the most obvious, that of a breaking wave. Perhaps the most iconic and enduring image is the superb print by Hokusai.

Katsushika Hokusai, The Great Wave of Kanagawa (1831), woodcut print, in "Thirty-six Views of Mount Fuji", Private collection. WikiArt.
Katsushika Hokusai, The Great Wave of Kanagawa (1831), woodcut print, in “Thirty-six Views of Mount Fuji”, Private collection. WikiArt.

More subtle and less well-known processes include the development (morphogenesis) of a foetus, which had defied detailed explanation for a long time. Between 1968 and 1972, René Thom (1923-2002), a French topologist, developed Catastrophe Theory to provide morphogenetic processes and other types of structural (in)stability with a sound basis. (I had the pleasure of attending a guest lecture by Thom at the Maison Française in Oxford; although given in French I still remember it as a highlight of my years there.)

Ocean waves have of course been well studied and do not need advanced topological concepts to predict when they are likely to break: that is a function of steepness. This analogy at least suggests that many of the ideas propounded already, such as the importance of ‘critical mass’ (as if you could weigh tweets or blog postings), may be erroneous. Instead one among many factors of importance in determining Twitterstorms may be the rate of rise in retweets, for instance. However this will need more detailed data.


There is already a substantial queue of organisations waiting to know how to create intensely viral events, and how to control them.

Words which appear in almost all writings about virality include ubiquity, connectivity, volatility, contagion, and a threat to capitalism (or whatever). Governments and large financial institutions therefore have obvious interests, mainly in controlling these events, lest they attack their own interests and stability.

Several of the research papers which have been published recently have come from joint US-Chinese teams, and Zhao et al. have in the last few weeks considered how there could be “immunization” to prevent the spread of rumours, because, as their abstract starts:
“Since most rumors are harmful, how to control the spread of such rumors is important.”

Commercial corporations are also investing to try to utilise viral techniques in advertising, although I suspect that most would not particularly wish to instigate intensely viral events for fear of losing control.

Assuming that this article is read by someone – always the blogger’s uncertainty – in the next I will look more closely at what happens in viral events.

References and further reading

Mackay C (various dates around 1841) Memoirs of Extraordinary Popular Delusions, Richard Bentley and various others. Available in original scans for volume 1, volume 2 and volume 3, and in the Project Gutenberg Library here. (Fascinating reading, although Mackay was more prolific than he was reliable, and many of his anecdotes have since been challenged.)
Nahon K & Hemsley J (2013) Going Viral, Polity Press. ISBN 978 0 7456 7548 0. (Detailed analyses of several viral events and discussion of social theory, but lacks any useful theory which could be used to predict or manipulate. Their emphasis is on ‘critical mass’ and speed of spread.)
Ni MY, Chan BHY, Leung GM, Lau EHY and Pang H (2014) Transmissibility of the Ice Bucket Challenge among globally influential celebrities: retrospective cohort study, British Medical Journal doi: 10.1136/bmj.g7185. (A straight epidemiological analysis of a near-viral event.)
Sampson TD (2012) Virality. Contagion Theory in the Age of Networks, University of Minnesota Press. ISBN 978 1 4529 3380 1. (If you enjoy reading Tarde and Deleuze, and think that you comprehend them, then you will probably enjoy this sociological analysis. Even the diagrams were a mystery to me.)
Standage T (1998) The Victorian Internet. The Remarkable Story of the Telegraph and the Nineteenth Century’s Online Pioneers, Weidenfeld & Nicolson. ISBN 978 0 297 841 487. (A brief and captivating account of early telecommunications, but does less to explore the social changes wrought by them.)
Zhao L, Wang J and Huang R (2015) Immunization against the spread of rumors in homogenous networks. PLoS ONE 10(5): e0124978. doi:10.1371/journal.pone.0124978. (Perhaps most interesting is the purpose behind this.)