On innovation

[I’ll continue the congenital beliefs series soon. Also chocolatarianist political theory] 

[Late note: Stefano Ghirlanda has pointed me at two papers he has coauthored on just this topic. In one, published in PNAS, he and his colleagues investigate a model of cultural innovation. In another, as yet unpublished, they argue that “creativity played a major role at the origin of human culture and for its accumulation throughout history, because cultural transmission cannot, on its own, generate exponentially increasing amounts of culture; 2) exponential increase in amount of culture can only occur if creativity is positively in?uenced by culture. They think creativity has increased. I need to work through the mathematical model to evaluate, and may one day comment.]

Many years ago I worked in public relations (the stain on my soul is almost gone now), and so when the world conference on public relations was held in my home town Melbourne, I decided to attend. Now PR hacks are pretty good at communication, practically, but when they got up to give “theoretical” accounts of what it was they did, it was just embarrassing, particularly the keynote speaker who tried to do a Shannonisation of PR. Basically, those who do, can’t theorise worth spit.

So a recent issue of the US National Endowment for the Arts magazine, entitled “What is innovation?” comes as no surprise. We get truisms personal anecdotes and not much else from the choreographers, artists, composers and performers who are interviewed, and it is no shame to them they aren’t philosophers. If they were philosophers, they probably wouldn’t be artists, choreographers, etc.

I have an interest in the topic, though. Apart from my recent series on the evolution of novelty (starting here), there is a real issue in cultural evolution and social evolution (distinct processes from biological evolution maybe) on how innovation occurs, and whether it has increased in frequency since the early stone age. Much is made, for example, of the shift from an early stone age technology (the Acheulean of Homo erectus) to a later one (the Mousterian for Neandertals and the Aurignacian for Homo sapiens) as an indicator of an increase in cognitive capacity.

What is innovation in cultural evolution, though? It is important to specify two kinds, or rather, two extreme poles of a continuum, of novelty. The one end is combinatorial novelty; something that is new in its structure or composition but not in its elements. Ideas that recombine older ideas in new ways, machinery that recombines existing technologies to do new things, ritual behaviours that recombine elements from other traditions to form new traditions are all examples of combinatorial novelty. Most of what artists do is combinatorial innovation.

The other end is what we could call deep innovation. This is where something entirely new, never done before, is tried and succeeds. The difference here with combinatorial novelty is that new elements are introduced into the mix. This is not entirely unlike Kuhn’s notion of “revolutionary” and “normal” science, for example.

There are two major problems with this account, which is due to Margaret Boden’s 1990 book, The Creative Mind. One is that we need to have an account of similarity that allows us to identify when something is “novel”. The other is that we need to be able to identify the right level of abstraction or description. As I argued in my series on evolutionary novelty, there are always deeper homologies: what looks like something new at one scale of description (the “phenotypic” or “morphological”; in actuality no such “levels” exist) is just a recombination of existing genes nearly all the time. However, not always. There are often new sequences of genes or gene regulators that did not exist (in that lineage) before and have arisen through mutation at the base pair scale.

However, the properties of a new gene are themselves the repurposing of both the physical (and therefore pre-existing) properties of the nucleotides, and the contextual properties assigned to those molecules by the wider system (DNA, no matter how novel, does little more than denature in a test tube, for example. It isn’t magic). So is this still a case of combinatorial novelty anyway?

It depends on the scale of the description. If you give a description at the genetic scale (i.e., washing out the information of the molecular properties and placing the nucleotides into a system context of a cell), then new genes from mutation are deeply novel. If you short the scale up or down, they may not be. Novelty is scale, and therefore description, relative.

This also gets us back to the question of similarity. Usually this is measured in terms of sameness of sequence in molecular biology, but it also has to be in terms of shape of the folded molecule and the shape of the interacting molecules elsewhere in the system. In culture, there is no simple metric to apply even like this one (which is not all that simple, it turns out). We might call this the “coding problem”. How do we code for cultural items in ways that don’t beg the question about similarity? This problem is long standing and probably not soluble, but I’d like to mention an index of similarity invented by Israeli psychologist Amos Tversky.

Tversky’s similarity measure (although some mathematicians don’t call it that for technical reasons) is basically the intersection of the shared properties of two things minus the unique properties of each thing (see this introduction). However, since the world doesn’t have a canonical list of properties, we have to define what Tversky called the feature set, or what others call the feature world. And choosing which things are to be counted is to game the system from the start. It’s not that item A and B are not similar in feature set F; it’s that they are similar in an infinitely large number of ways in an infinitely large number of Fs, each one of which is objective.

So measuring the rate of innovation in cultural evolution is highly contextual and dependent on how the features are chosen and described. We should be very careful of claims that some technology, art or period is more or less innovative than another without a clear specification and justification of the Fs chosen.

Let us ask whether the rate of innovation is different now than, say, 50,000 years ago. The presumption is that we are much smarter, better innovators, and more inventive than our forebears. Sometimes we explain this in terms of native abilities – we have better machinery than they did. In one sense, at least, this might be true. We do things none of our distant forebears could do. But whether that applies over such a short period, evolutionarily speaking, as 50ky, is highly questionable.

Cultural evolution appears to me to have a constant rate. There is an a priori reason for thinking this. Humans are not, so far as we can tell, very different across time or space for some time now. As Jared Diamond once observed, in a single generation a stone age technology culture can go to flying jet aircraft. But the innovation of the jet aircraft is not so rapid or easy. Are we that much smarter? Maybe it’s to do with the economic and cultural system itself. And that also has to have some purchase: our culture can do things the Romans couldn’t. We have manufacturing industry, scientific research systems, and education.

But none of this seems enough to justify the claim that we are producing more novelty per capita than our forebears or distant cousins. Suppose that it takes a certain period to learn and master a technology, either to make it or use it or both. Suppose there is a certain error rate in transmission of these skills, so that unless somebody refines and corrects the technique from experience it will become less well adapted to its purpose. Suppose that the people who have the capacity and opportunity to do this occur at a flat rate in the population. From this we can infer that a population with a limit to the size and degree of communication, and which is too small to produce these innovators, will lose technologies at a constant rate.

If one of these technologies (say, growing grains or making cheese) permits a population to grow in size, the rate of retention will increase until a threshold is reached at which new techniques that work (just by chance) will tend to be preserved as well as existing technologies. The absolute rate of innovation will rise for that population simply as an effect of increasing population size, even though the relative or per capita rate is constant. And the growth curve for innovation will initially be exponential, until it approaches the resource limit for innovating, employing and transmitting the information, at which point we have a standard logistics curve:

So we can get some pretty rapid innovation without needing to posit that the rate of innovation has increased at all. In the absence of any evidence that we are somehow smarter than our paleolithic ancestors, I think I’ll stick with this account as sufficient.

So we have the following situation: (1) we cannot objectively identify novelty; (2) we cannot objectively set a scale of description that will allow us to measure novelty, and (3) we do not need to posit an increase in innovation beyond the number of people in a population and some constraints on communication. Objections to evolution on the basis that there is no “source of novelty” (sometimes “source of new information” in intelligent design circles) rely on the falsehood that there is something necessarily objective in the way we describe the world. Actually, we describe what looks of interest to us, and we get surprised that sometimes things are not captured by our representations.

Boden, Margaret A. 1990. The creative mind: myths and mechanisms. London: Weidenfeld and Nicolson.

Tversky, Amos, and Eldar Shafir. 2004. Preference, belief, and similarity: selected writings. Cambridge, MA: MIT Press.