Not only do human beings have no vision of the future as it varies from the present, but often a very imperfect understanding of the present in front of their eyes. This was demonstrated by an experiment (Derex, Bonnefon, Boyd and Mesoudi (DOI 10.17605/OSF.IO/GE7CS, not yet published) which Maxime Derex presented at the Tartu conference 2018. The following is from the abstract and from memory.
In the experiment the initial participants were presented with a ramp, a hub with four spokes at the cardinal points of the compass, and four weights one of which could be placed at a number of determinate points on each spoke. It was a transmission chain experiment, and the task was to determine the optimum placing of the four weights for the shortest possible rolling time between the top and a set point on the ramp. Each initial participant was given five trials, and the results of the last two, both the placing of the weights and the average speed, were passed on to the next participant in the chain, which was five participants long.
By the end of the chain the time taken for the centre of the hub to pass the finishing line had decreased, tending towards the optimum.
In a second version of the experiment, the participants were allowed to pass on not only the data for their last two results, but also their hypothesis about the optimum placing of the weights.
The additional information, the hypotheses, made no significant difference to the outcome. In fact Maxime said that a) some of the hypotheses were very weird indeed, and that b) it was not formally possible to work out a solution without knowing the diameter of the hub and the distance between start and finish on the ramp.
So no participant had a formally coherent, verbally expressible hypothesis about the system that was of use to the next person in the chain. And yet along the chain performance improved towards the optimum.
The title of the paper was Causal understanding is not necessary for the improvement of culturally evolving technology, and the experimenters were focussed on establishing whether this was true.
Clearly the participants had to have had some degree of causal understanding of the system, otherwise their choices would have been entirely haphazard. They might have understood that the weight placed at the outer end of a lever moving downwards will have more axial effect than one placed at the inner, hub end of the lever; while not understanding that placing a weight at the outer end of the lever increases the moment of inertia of the whole wheel, which has a negative effect on its acceleration. So what the participants lacked was an exhaustive, coherent and unified theoretical understanding, such that could produce by means of calculation the optimum result without physical trial. They had partial, particulate understanding, but not total understanding.
The startling result of this experiment is that it demonstrates in a clean, uncluttered setting that the solution to problems can evolve where two dimensions interact; the accumulation of particulate causal understanding, and feedback, that is to say selection, from the proximate environment, in this case the experimental apparatus. The reciprocal of this relationship is also the case, but that is to over-complicate things at this point.
To generalise, human collectives (here modelled in the transmission chain) can have a distributed particulate understanding of a system. The serial nature of the experiment, first one participant and then the next, delivers the misleading impression that the emergence of the near-optimum result is achieved in a series of, admittedly cryptic, logical steps which themselves are based on the logic inherent in the partial and particulate knowledge of causation in the transmission chain. But this would be to suppose that by chance each chain of five participants was ordered 12345 in a sequence whereby the small discrete logical sequences of their particulate understanding met as it were end to end in a superordinate logical sequence; one set of axioms and their conclusion, though only part of the solution, leading neatly to the next. This is so unlikely as to be hardly worth testing, which is lucky, because to test it you would have to run each chain 120 times in all their possible sequences, inducing in each participant a precise and total amnesia about the previous sequence before starting the next.
It seems that, though time is a necessary dimension in which the experiment took place, the functional causal understanding is distributed in a way indifferent to time sequence, as it were in a three dimensional space in which the direction of time is irrelevant as far as running order, and therefore order of bits of understanding, is concerned, but crucial for the sequence of instances of selection by the apparatus.
Put at its most plonkingly obvious, for the solutions to move towards the optimum, it is not the causal understanding of any individual brain which is necessary for this tendency, but the causal understanding of the collective of the five members of the chain, as selected by the apparatus.
For something to be selected upon, there must be discrete variation, as with alleles on the chromosome. At the particulate level of analysis, at whatever scale one can posit alternatives, the participants must be choosing between one case or its alternative, the cases or their alternatives being possibly arranged in multi-dimensional interactive hierarchies. This is the kind of thing the human brain does, and is extremely good at. The fact that the experimenters themselves do not think it worthy of note demonstrates that the process is almost totally opaque to the functions of the brain that constitute what we each refer to (in English) as I or me. I will come to this “self” in time. Meanwhile the objective situation remains, that neither the participants nor the experimenters had any idea of what the process of selection was, nor what at the particulate scale the alternative units of selection are. So that’s the big question: in the evolution of human culture, what are the discrete objects constituting the collective upon which selection is made. It’s another way of saying, we are in search of a replicator.
That is the huge value of the Derex et al. experiment, that it tacitly frames that question with relentless inevitability.
That inevitability is born out “in the field”. Chris Buckley’s and Eric Boudot’s The Evolution of an Ancient Culture (Buckley, 2017) describes and examines the evolution of looms in East and Southeast Asia. The scope and depth of this study is to me awe-inspiring. I hope it will be seen as one of the great landmarks in the study of the evolution of the material culture of Homo sapiens. I’ll quote [in italics] the parts that are most relevant to our purpose, and comment on them, but these quotes give little indication of the range and complexity of the cladistic and phylogenetic analysis.
It is not necessary for a novice weaver to have a general understanding of how the loom works (a ‘theory of the loom’) in order to begin weaving.
This makes the point I have already made with the apprentice brick maker, that you learn by doing, not by theory.
In the case of the complex frame looms and patterning devices used in [Southwest China], weavers are unable to recreate these devices from memory, and only a small proportion of talented weavers appear to understand in detail how they work. A carpenter can help a weaver to make a copy of a loom, but only if an existing model is available.
This emphasises the amount of information inherent in the object, the thing-in-the-world. I guess demonstration of “understanding in detail how they work” was dependent on the presence of the loom, and was of the “this does this which leads to that” basis, with hand gestures the eye movements of the explained-to could follow; and not a purely verbal account in the absence of a loom. “Weavers are unable to recreate these devices from memory” substantiates the general point, that all our knowledge of our material culture depends on its persistent presence in the world, and if it were all to disappear on the instant, our “knowledge” would vanish with it.
For example, in one village in Guangxi where a particularly complex loom with a pattern-recording system is employed, only one weaver could be found who had the skill necessary to construct a patterning system for a new motif . It appears that the level of skill needed to modify or renew a patterning system has always been relatively rare, with most households relying on hand-me-down pattern systems that have been in use for several generations.
My knowledge of weaving goes little further than the enormous-pixel particulate understanding of “there is a warp and there is a weft.”
However, as (Buckley, 2017)’s illustration shows, evolved looms are a lot more sophisticated than that. The patterns woven are beautiful as well as complex, and are recorded in the toroidal “drum” hanging from a beam at the top. I have absolutely no idea http://rsos.royalsocietypublishing.org/content/4/5/170208how this works. What can be advanced from the evidence as a first assumption is that if all these drums, the physical loci of the pattern-making system, were simultaneously lost, they could not be recovered from particulate understanding of the system remaining in the collective of human brains. They would be extinct. The most that it appears the particulate understanding of a rare expert can achieve is “to [exercise] the skill necessary to construct a patterning system for a new motif”. It would be of huge interest to know how new drums do emerge, which they evidently do, or maybe at least did before being swamped by industrialisation. It is possible, for instance, that none was ever constructed all at once and as a whole. The armature around which they are constructed is presumably a simple given and relatively easy to assemble, whereupon patterns might be added one at a time, by the above experts, whose existence was crucial; as and when required, until the drum became full or worn out; and each drum was possibly unique in its particulate configuration. These developing pattern drums contained and could transmit the particulate understanding, how to construct the patterning system for a new motif, which was, with the armature, all that was necessary for the emergence of a new drum.
The outstanding characteristics of the transmission processes are a lengthy apprenticeship that encourages ‘over-learning’, the orientation of older weavers towards detecting and correcting errors (deviations from tradition practice) and the codification of complex tasks into ritualized procedures. All of these features tend to increase the fidelity of transmission and discourage innovation.
Here we have the ground state necessary for evolution, replication with fidelity over time, with the margin of minute variation which is available to selection kept to a minimum. The dangers of “innovation” and “invention” are apparent. If an autocrat dictated that all present pattern drums were to be destroyed by fire and replaced with a model invented by one of his favourites, and it was discovered a little way down the line that the new model had severe limitations, in fact was rubbish, the old pattern drums would be unrecoverable. Thus
Our analysis of the transmission processes of traditional weaving cultures, however, shows that actual behaviours are mostly concerned with reducing errors and discouraging innovation.
In both the Derex, Bonnefon, Boyd and Mesoudi experiment, and Buckley and Boudot’s paper, two big rivers meet; data from things-in-the-world, and a process of selection upon particulate information which is replicated with fidelity but has an envelope of fractional variation; data and Darwin.
The task still remains, to identify the replicator.