On the productive use of randomised behaviour

It is boredom that drives Luke Rhinehart (Rhinehart, 1971) to act on the outcome of dice rolls in The Dice Man. He assigns actions to each side of a die which he then casts. His behaviour therefore is randomised in so far as it is not predictable in advance.
The cultural impact of The Dice Man was largely due to the shockingly violent and immoral behaviour in which he engaged with inadequate justification. The randomness he has purposely introduced gives his action an insupportability. It is, from one perspective, de facto meaningless and incomprehensible. This is shocking because this is not ‘normal’. As with any senseless act which is often the product of a crazed mind, it is our inability to ascribe or make sense of it, that drives our perturbance.
Part of what is unsettling about The Dice Man is the alien nature of the decision-making process Rhinehart employs. It is counter to our rationality to make choices on the basis of pure chance. This is not how humans work; we think things through and act for sensible and logical reasons. But do we and should we?

Putting evolution to work
Evolution has been described as design without a designer. An intelligent God would simply ‘build’ persons ready-made to a template. But evolution has no blueprint or end-product in mind as it ‘designs’. It is a process of trial and error; each model that rolls off the production line is tested to destruction and its offspring (if any) are partly modified by random means.
A child will inherit certain characteristics from its forebears but in a manner that is deliberately randomised. The nature of the randomness (presumably itself the consequence of an evolutionary process) rewards and preserves certain qualities but allows scope enough for new and dramatically divergent qualities to emerge. In other words, our offspring will inherit a great many of our qualitites in like proportions, but may still exhibit characteristics that are harder to simply attribute to either parent.
This mechanism is deliberately employed by Nature as an efficient search strategy when seeking a ‘well-designed’ person ie a person fit for the purpose of surviving in their environment. And the employment of randomness in this regard is key.

Global and local, minima and maxima
Imagine trying to locate the highest mountain in a range that one can only explore on foot. To do this, one might employ a simple strategy such as,
Strategy 1: “I shall walk forward five steps and review. If I have ascended at all, I shall repeat the strategy. If I have not ascended or descended, I shall turn to face left and repeat the strategy. If I have descended, I shall turn 180 degrees and repeat the strategy.”
Something like this, might lead one up the summit of a high local peak. It will effectively interpret any ascencion as good news. If we move away from the local peak, we turn around and head back up. Now this kind of search strategy (which is deterministic and not in any way randomised) may be successful at finding local peaks but as soon as a local summit is reached, it marks an end to our search; thereafter an movement away from the summit is immediately reversed as though we are tethered to the apex.
This strategy, I contend, is roughly approximate to how people ‘rationally’ live. With a goal in mind, we employ a reasonable strategy that we can test along the way, a strategy that in advance seems likely to lead to some sort of success. This strategy will, deductively, lead to a local maxima and turn away from local minima.
But for an Evolutionary designer, this would not be enough. Because this strategy is likely to be sub-optimal in the long term. After all, we can be certain that this strategy will be an evolutionary dead-end. Once we reach a summit of any sort, we have no means of escape and a potentially better outcome can never be achieved.
In contrast, evolution allows such strategies to be subject to random alteration. So imagine now that the strategy above becomes;
Strategy 2: “I shall walk forward five steps and review. If I have ascended at all, I shall repeat the strategy. If I have not ascended or descended, I shall turn to face left and repeat the strategy. If I have descended, I shall continue for a random number of steps up to a total of five, turn about a random number of degrees and then repeat the strategy.”
This strategy still favours ascent. But it no longer sees descent as an evil to be avoided at all costs. The introduction of random behaviour allows for an infinite array of outcomes; this strategy will undoubtedly reach diffenent, and higher, local maxima, in time and probably reach the global maximum. It may forgo notable local maxima in the short term and spend long periods in local minima. But the possibility of better outcomes is now available and the evolutionary dead-end has been avoided.

Using random behaviour to maximise outcomes
It is striking how little we allow our behaviour to follow even a mildly random path. If we intend to deploy a new marketing strategy, it will typically be planned in every detail with nothing left (literally) to chance. And yet we could employ evolutionary techniques by allowing our strategies to adapt and modify randomly. This might mean altering a ‘successful’ strategy for the worse, knowingly. And herein lies the rub; this creates a tension with our rational faculties. Why change a winning formula especially when we can reason that the change will be detrimental? And why persist with a losing formula that has no obvious reasonable chance of success? The justification for so doing, is that by analogy, this is the mechanism employed by Nature for successful evolutionary design. Randomising variation between generation means that winning formulae can indeed be reversed and losing formulae made persistent. But this is the route to global maxima and to avoiding local dead-ends.
The question then really pertains to the appropriateness of the analogy. Our marketing strategy may only be expected to run for a few months in one or two locations. It would be undoubtedly foolish to choose a strategy that our cognitive faculties suggest will fail in all circumstances in this time and in these places. But if the backdrop is amenable, the duration is long enough, the number of instances adequate, then randomising the strategy to a certain extent is not merely preferable, but is likely to be optimal. Because Nature says so.

Advertisements

Leave a comment

Filed under Lateral thoughts