Alex Bentley is Chair of Anthropology at University of Tennessee, Knoxville. His new book with michael O'brien is: The Acceleration of Cultural Change: From ancestors to algorithms (MIT PRess, 2017)

Chance and evolution

Some years ago, the New Scientist  ran a cover with a cartoon drawing of a wide range of cuddly animals (pandas, lemurs, grizzlies, tigers) animals lying on an around an enormous red die.  The title was “Accidental origins: The idea that completes Darwin’s evolution.” The story, about the role of chance in evolution, recalled what Stephen J. Gould had famously suggested, decades ago, in that if we “replayed the tape” of evolution, much of the life that evolved on earth could be completely different. While the general evolutionary process would be the same, random, chance events in the dawn of species would happen differently before selection had grabbed hold to sort and refine the species. 

Though often neglected, the role of chance is inherent to evolution. Fifty years ago, when Motoo Kimura and James Crow introduced the Neutral Theory of evolution, they essentially asked: what would happen if there were no selection, if things just replicated and change came about by chance? 

As Lewis Spurgin discussed in Aeon, chance and luck are as much the essence of evolution as determined selection. The New Scientist cover was referring to a Nature article that had examined something very simple – the distribution of branch lengths in the evolutionary trees (phylogenies) of different species, inferred from the fossil record. The researchers had tallied all the branch lengths and found that, rather than a bell-curves, three-quarters of the phylogenetic trees showed and exponential distribution of lengths -- a steeply descending function by which most were very short. With this simple result, the researchers concluded that much of species evolution has been stochastic: driven by random events, as opposed to a an adaptive process guided by selection. 

There are lessons here for cultural evolution, but first, can we really gain insight by tallying up the evolutionary branch lengths?  Absolutely; there is information in a distribution of events, and many, many data sets -- from sales records to social media analyses to historical lists -- contain records of events. The classic bell-curve, which we assume for so many things, from IQ to the sizes of bicycle tyres, has merely been a useful assumption often for equilibrium systems (the Central Limit Theorem). As a non-equilibrium process, however, evolution yields different distributions, with different implications.  The Poisson distribution, for example, indicates that change is driven by random events.

On the geological scale of species evolution, random events might include volcanoes or meteors. Success on social media is also driven by chance, which is "why predicting your social content will `go viral' is stupid."

Even science, as Lewis Spurgin pointed out, is driven by accidents and chance:

"We constantly make subjective decisions as scientists: which questions get us fired up, which do we ignore, when do we consider a result significant enough to publish, how do we approach an analysis, and how do we interpret our findings." (Spurgin, Aeon, 09 July, 2013).

Ideally, we'd like science to build knowledge cumulatively, over generations. The same applies for cultural knowledge. We'd rather these knowledge systems not just build by chance, so what steps can be taken? Mike O'Brien and I discuss this in our new book with M.I.T. Press.

A little bot goes a long way: randomness and collective coordination

The Acceleration of Cultural Change: From Ancestors to Algorithms

The Acceleration of Cultural Change: From Ancestors to Algorithms