Memes, Genes and Evolution on Facebook

The way memes evolve on Facebook is startlingly similar to the way genes evolve on Earth, say researchers who have carried out the first large-scale study of memetic evolution 


One of the more extraordinary ideas about life is that it is fundamentally a phenomenon of information. The thinking is that the basic ingredients for the kind of complex systems we call life are not water or carbon or even DNA. All that’s needed is information and the process of evolution.

By that way of thinking life is substrate independent. It can just as easily evolve in silico in a computer simulation as it can in carbon-rich soups on Earth. Indeed, numerous researchers have succeeded in creating what look like simple life forms in computer simulations of evolution (although complex life has so far proved elusive).

But it is not just in simulations that information evolves. Back in 1976, the evolutionary biologist Richard Dawkins introduced the idea of a meme in his book The Selfish Gene.

Dawkins’ suggestion was that ideas, behaviours and styles can spread from person to person within a culture, changing as they jump. And his key insight was that this kind of change is a form of evolution that is analogous to the same process that influences genes.

So memes ought to evolve in exactly the same way as genes: by a repeated process of selection, reproduction and mutation.

But while Dawkins’ idea is interesting and increasingly popular, how close is the analogy between memes and genes? Do memes really evolve in the same way as biological entities, which follow precise mathematical laws in terms of their rates of mutation and so on?

Today we get an answer thanks to the work of Lada Adamic affiliated to both Facebook and the University of Michigan in Ann Arbor and a few pals. For the first time, these guys have measured how memes spread through a big social network (Facebook), how they reproduce and how they mutate along the way.

Their conclusion is fascinating: it turns out that memes do indeed spread, mutate and evolve in ways that are mathematically identical to genes. However, there are important differences too. And these guys say that understanding this process can give deep insights into the way information spreads through cultures and the way individuals change it as it spreads.

What makes this kind of work possible for the first time is the huge database of information gathered by Facebook between April 2009 and October 2011. This consists of thousands of memes that have together appeared over 460 million times.

The first task for Adamic and co was to analyse these memes, to find out how they spread and how they changed during this process.

A good example is the meme: “No one should die because they cannot afford health care and no one should go broke because they get sick. If you agree please post this as your status for the rest of the day.”

This began life as a political meme supporting President Obama’s health care reform bill, and appealed mainly to liberal-leaning individuals, a fact Adamic and co were able to confirm by calculating the average political leaning of people who both filled out this field on Facebook and who posted this meme. This meme was copied some 470,000 times.

However, the meme quickly mutated. A version that included the phrase “[Your Name] thinks that…” appeared 60,000 times. Some variants used the phrase “the next 24 hours” rather than “rest of the day”. The third most popular variant included the phrase: “We are only as strong as the weak among us”.

Humourous version also appeared such as: “No one should be without beer because they cannot afford one…”. And others appeared that seemed to criticise the bill and would therefore be popular among conservatives, for example: “No one should die because the government is involved in healthcare….”

Adamic and co analysed this and other memes by working out the percentage of copies that included a change from the original and also calculated the degree of change by counting the number of characters that changed between variants. “The first striking pattern in the data is the uneven popularity of the variants,” they say.

What’s striking about it is that the pattern is almost identical to biological distributions, such as the number of species per genus, bacterial populations and protein domain family sizes. This kind of distribution comes about because of a process known as the Yule process (or preferential attachment) and is the result of nothing more than replication and mutation.

“The Yule model is a good fit for the great majority memes which are ideas evolving “naturally” under low mutation rates,” they say. These memes evolve in exactly the same way as genes.

And there are other similarities too. For example, there is a small but clear preference for variants that are shorter than the original memes. That’s analogous to bacteria favouring small genomes because it allows fast replication. “Both memes and genes are shaped by the same two factors: their length has to be sufficient to encompass information that helps them to replicate, but excessive length slows down replication,” say Adamic and co.

Another interesting similarity is that the same advantageous sequences can appear in many different memes, probably transferred by a single individual from one meme to another. “This process is analogous to lateral gene transfer in bacteria wherein, useful genetic code, e.g. plasmids conferring antibiotic resistance, can be transferred through mechanisms other than replication,” say the team.

And then there is meme fusion, where two memes combine into a single one. Exactly this process also happens in genomes where two single genes combine to create a single gene that retains the function of both. “We observed several fusion events between different memes, some giving rise to popular variants,” they say.

Of course, Adamic and co also found important differences between the behaviour of memes and genes. Perhaps most significant is that evolution is a blind process in biology but not in social media where there can be a conscious effort to create mutations that will spread more effectively.

That leads to some memes evolving with very high replication rates that the Yule process cannot account for. That comes about because many memes simply ask that they are copied and pasted into a person’s status box but some ask for additional information.

For example, one meme asks for for the birth weights of children, another for oldest friends and yet another for billboard charts from birth days and so on. Clearly these will have higher rates of mutation.

Nevertheless, the discovery that memes evolve in ways that are amazingly similar to genes is hugely interesting. It is a far cry from the evolution of more complex objects or indeed life itself but it is interesting nonetheless.

The blind process of evolution has produced the most extraordinarily complex devices, such as the human brain. But it has taken many billions of years to achieve this.

It’s not clear that anything like this will ever be possible with memes. But there’s no harm in dreaming.

One thing though: the evolution of memes is a much younger process but it’s easy to imagine the potential. The missing ingredient so far? Time.

Ref: arxiv.org/abs/1402.6792 : Information Evolution in Social Networks

Follow the Physics arXiv Blog by hitting the Follow button below, on Twitter at @arxivblog and now also on Facebook