Adaptive Capacitytaxonomy/feature.php

Eye on the target - Identifying what is being adapted for:

All complex systems involve channeling flows in the most efficient way possible - achieving the maximum gain for the minimal output - and 'discovering' this efficiency can be thought of as achieving a 'fit' behavior. When looking at a system's adaptive behavior, one therefore needs to first understand how fit regimes are operationalized, by considering:

1. What constitutes a 'fit' outcome;
2. How the system registers behaviors that move closer to this outcome (see Stigmergy end handlebar);
3. The capacity of agents in the system to adjust their behaviors so as to better align with strategies moving closer to the 'fit' goal.

It is this third point, point pertaining to the 'adaptive capacity' of agents that we wish to look at in more depth.

Variation, Selection, Retention (VSR):

If we consider the example of ant trail formation, behaviors that lead to the discovery of food are more 'fit'. Adopting the lens of Variation, Selection and Retention, the system unfolds as follows:

1. A collection of agents (ants), seek food (environmental differential) following random trajectories (Variation).
2. Ants that randomly stumble upon food leave a pheromone signal in the vicinity. This pheromone signal indicates to other ants that certain trajectories within their random search are more viable then others (Selection).
3. Ants adjust their random trajectories according to the pheromone traces, reinforcing successful food pathways and broadcasting these to surrounding members of the colony (Retention).

What emerges from this adaptive process is an ant trail: a self-organizing phenomena that has been steered by the adaptive dynamics of the system seeking to minimize the global system energy expended in finding food. What is important to note is that the adaptation occurs at the level of the entire group, or system. The colony as a whole coordinates their behavior to achieve overall fitness, with food availability (the source of fitness) being the differential input that drives the system. The ants help steer one another and, overall, the behavior of the colony is adaptive. Individual ants might still veer off track and deplete energy looking for food, but this remains a useful strategy in cases where existing food sources become depleted. Transfer of information about successful strategies is critical to ensuring that more effective variants of behavior propagate throughout the colony.

None of this is meant to imply that, if the ants follow this protocol, they will find the most abundant food source available. Complexity does not necessarily result in perfect emergent outcomes. What it does result in is outcomes that are 'satisficing' and that allocate system resources as effectively as possible within the constraints of limited knowledge. Further, the system can change over time, meaning that other, more optimum performance levels may be discovered as time unfolds.

Capacity to Change:

An agent's ability to vary its behavior, select for behaviors that bring it closer to a goal, and then retain (or reproduce), these behaviors, is what makes agents in a complex system 'adaptive'. If agents do not possess the capacity to change their outputs in response to environmental inputs, then no adaptive processes can occur.

While this might at first seem self-evident, this basic concept can often be overlooked. In particular, it is easy to think about a system composed of diverse components as being 'complex' without considering whether or not the elements within the system have some inherent ability to adjust in relation to this complex context -

Example:

Consider an airplane. It is a system comprised of a host of components and together these components interact in ways that makes flight possible. That said, each component is not imbued with the inherent ability to adjust its behavior in response to shifting environmental inputs. The range of behaviors available to the plane's components are fixed according to pre-determined design specifications. The machine components are not intended to learn how to fly better (adjusting how they operate) in response to feedback they receive over the course of every flight.

If we try to understand an airplane as a complex system, and identify 'flying better' (using less energy to go further) as our measure of fitness, then would it be meaningful to speak about the system adapting? It the agents in the plane's system are the individual components, are they capable of variation, selection, and retention? Even if we were to model system behavior from the top down testing design variants in components, the system itself would not really be 'self-organizing' in this instance: without external tinkering nothing would happen.

'Seeking' fitness without volition:

Does it follow that inanimate objects are incapable of self-organization without top down control?

It is reasonably easy to understand adaptation within a system where the agents posess some form of volition. What is intriguing is that many complex systems move towards fit regimes, regardless of whether or not the agents of the system have any sort of 'agency' or awareness regarding what they do or do not do.

Example: Coordination of Metronomes:

In the video below, we see a group of metronomes gradually coordinating their behaviors so as to synchronize to a regular rhythm and direction of motion. While this is an emergent outcome, it is initially unclear how one might see this as a kind of 'adaptation'. But if we look to the principles of VSR, we see how this occurs. First we observe a series of agents (metronomes), displaying a high degree of variety in how they beat (in relation to one another). The system has a shared environmental context (the plank upon which the metronomes sit), which acts as a subtle means of signal transfer between the metronomes. The plank moves parallel to the direction of metronome motion, creating resistance or 'drag' in relation to the oscillation of the metronomes on its surface. Some individual metronome movements encounter more resistance in relation to this environment (the sliding plank), while some movements encounter less (a more efficient use of energy). These differentials cause each metronome to alter its rhythm - ever so slightly - until all metronomes move in sync.

Watch the metronomes go into sync!

Considered as VRS we observe the following:

1. There is a Variation in the metronome movements with certain oscillatory trajectories encountering more friction and resistance then others;
2. The physics of these resistance forces create a Selection mechanism, whereby each metronome alters its oscillatory patterns in response to how much resistance it encounters.
3. As more metronomes enter into coordinated oscillating regimes, this in turn generates more resistance force being exerted on any outliers, gradually pushing them into sync. Once tuned to this synchronized behavior,  the system as a whole optimizes its energy expenditure, and the behavior persists (Retention).

Keep it to a minimum!:

The system adapts to the point where overall resistance to motion is minimized. The metronomes 'achieve' the most for the least effort: a kind of fitness within their context.

While the form of 'minimization' varies, all complex systems involve seeking out behaviors that conserve energy - where the system, as a whole,  processes the flows it is encountering using the least possible redundant energy. While this cannot always be perfectly achieved, it is this minimizing trajectory that helps steer the system dynamics.