Bottom-up Agents

What might an Agent 'B'?

Complex systems are composed of populations of independent entities that nonetheless form a particular 'class' of entities sharing common features. Agents might be ants, or stocks, or websites. Furthermore, they might be Bikes, Barber shops, Beer glasses, or Benches (what I will refer to  below as the 'B' list). We can ask what an agent is but we could equally ask what an agent is not!

Defining an agent is not so much about focusing on a particular kind of entity, but instead about defining a particular kind performance within a given system and within that system's context. Many elements of day-to-day life might be thought of agents, but to do so, we need to first ask how agency is operationalized.

Example:

Imagine that I have a collection of 1000 bicycles that I wish to make available for rent across a city. Could I conceive of a self-organizing system where bikes are agents - where the best bike distributions and locations emerge, with bikes helping each other 'learn' where the best flow of consumers is? If a bike's 'destiny' is to be ridden as much as possible, and some rental locations are more likely to enable bikes to fulfill this destiny than others, how could the bikes distribute themselves to as to maximize fulfillment of their collective destiny?

What if I have 50 barber-shops in a town of 500 000 inhabitants - should the shops be placed in a row next to one another? Placed equidistant apart? Distributed in clusters of varying sizes and distances apart (maybe following power laws?). Might the barber shops be conceptualized as agents competing for flows of customers in a civic context, and trying to maximize gains while learning from their competitors?

And what about beer glasses: if I have a street festival where I want all beer glasses to wind up being recycled and not littering the ground, what mechanisms would I need to put into place in order to encourage the beer glasses to act as agents - who are more 'fit' if they find their ways into recycling depots? How could I operationalize the beer glasses so that they co-opt their consumers to assist in ensuring that this occurs?. What would a 'fit' beer glass be like in this case (hint: high priced deposit?). 

Finally, who is to say where the best place is to put a park bench? If a bench is an agent, and 100 benches in a park are a system, could benches self-organize to position themselves where they are most 'fit'?

The examples above are somewhat fanciful but they are being used to illustrate a point: there is no inherent constraint on the kinds of entities we might position as agents within a complex system. Instead, we need to look at how we frame the system, and do so in ways where entities can be operationalized as agents.

Operational Characteristics:

The agents above can each move into more fit behavioral regimes provided that certain operational dynamics are in place: 

• having a common Fitness criteria shared amongst agents (with some performances being better than others),
• having an ability to exchange Information amongst other agents, which helps direct and constrain how each agent behaves  (get to better performance faster).
• having an ability to shift performance, or Adaptive Capacity  (see also Requisite Variety),
• operating in an environment where there is a meaningful difference available that drives behavior (see Driving Flows)

Thought Experiment:

Let's take just one of the examples above. The location of bikes (you can also find another example of the park benches on the Governing Features page text.

Let's begin by co-opting a number of parking spaces in a city as temporary bike rental stations. Bikes are affixed to a small rolling platform in a vacant parking stall that holds 4 locked bikes.  These bike stations are then distributed, at random around a neighborhood. Individuals subscribe to a service that allows them to use bikes in exchange for money or bike credits.

• Let us assume that the ultimate 'destiny' of a bike is to be ridden. Then the frequency at which this destiny is manifested would be considered its measure of fitness. For purposes of this thought experiment lets assume that each bike can measure this fitness: it has a sensor that detects ridership.
• Let us then assume that each bike station is equipped to receive signals from the bike stations in its vicinity, indicating if bikes at those stations are being borrowed or not. With this information a bike station can calibrate which of its nearest neighbors are most readily fulfilling their destiny of being utilized.
  
• Let us then assume that the bike platforms are given a bit of 'smart' functionality - they are connected to an app, that those subscribing to the rental service have on their phone. If a bike station is under-performing in comparison to its neighbors, it will offer a credit to any user of the service who will hitch up a bike to the rolling bike station, and move it to the nearest location of higher use.  This gives the bike stations the ability to shift location, providing adaptive capacity.
• Finally let us assume that enough people are using the app, such that variations in use frequency provide enough data to mark trends or be useful. These usage flows then mark trends within the bike rental system, with certain bike station locations being popular, others not so.  As people rent or do not rent bikes, a source of difference enters the system, with certain bikes receiving more or less flows of users

It should be rather intuitive to image what would happen in this system. Some bike stations will capture more flows of people than others - the reasons for this might not be clear, and may vary from day to day depending on different conditions.  The reasons do not necessarily matter. From the perspective of the bike stations (as the agent in the system) the reason why a particular location is better or worse is not important, what matters is that bikes that are underutilized will gradually readjust their position in the city so as to better capture the flows they crave. Overtime, sites that have a high usage demand will achieve consolidations of bike stations, with each station adjusting its position based on information gathered from its nearest neighbors. This will continue until such time as all stations are positioned in ways where they are all capturing an equal number of usage flows, with none able to move to a better location. A kind of system equilibrium has been reached. Other equilibrium states may also exist, and so it is helpful if bike stations occasionally abandon this stable state, to randomly explore other potential, unoccupied sites that may in fact harbor unharnessed flows of bike ridership. It should be noted that the density of the emerging bike hubs can vary dramatically. There may be areas where 10 stations, 20 or only 1 station is viable. The point is that the agents in the system can distribute themselves, over time, to service this differential need without need for top down control. Here we have an example of a kind of 'swarm' urbanism.

This example is not typical of those given in complex adaptive systems theory, but it helps illustrate how it is possible, at the most basic level,  to conceptual a systems of complex unfolding by using only the notions of Agents, Fitness, Adaptive Capacity, Driving Flows and Information. There are other more nuances, but any of the systems listed above (the bicyles, barber shops, or beer glasses), could be made to function using the same basic strategies. 

'Classic' Agents

The list of potential agentic entities offered above - the 'B' list - is somewhat odd.  We begin with them so as to avoid limiting the scope of what may or may not be an agent. That said, this collection of potential agents are not part of what might be thought of as the 'canonic' Agent examples  - what we might call the 'A' list -  within complexity theory. Let us turn to these now:

Those drawn to the study of complex systems are often compelled to explore agent dynamics because of certain examples that demonstrate highly unexpected emergent aspects. These include 'the classics' (described elsewhere on this website) such as:  emergent ant trails, coordinated by individual ants, emergent percolation patterns, coordinate by water molecules in Benard/Rayleigh convection, emergent higher thought processes, coordinated by individual neurons firing.

In each case, we see natural systems composed of a multitude of entities (agents) that, without any level of higher control, are able to work together to coalesce into something that has characteristics that go above and beyond the properties of the individual agents. But if we consider the operational characteristics at play, they are no different from the more counter-intuitive examples listed above. Take ants as an example. They are an agent that has:

• a common fitness criteria shared amongst agents (getting food),
• the adaptive capacity to shift performance (searching a different place)
• an ability to exchange information amongst other agents (deploying/detecting pheromones)
• an environment where there is a meaningful difference that drives behavior (presence of food sources/flows)

Ant trails emerge as a result of ant interaction, but the agents in the system are not actively striving to achieve any predetermined  'global' structure or pattern: they are simply behaving in ways that involve an optimization of their own performance within a given context, with that context including the signals or information gleaned form other agents pursuinng similar performance goals. Since all agents pursue identical goals, coordination amongst agents leads to a faster discovery of fit performance regimes. What is unexpected is that, taken as a collective, the coordinated regime has global, novel features. This is the case in ALL complex systems, regardless of the kinds of agents involved.

Finally, once emergent states appear, they constrain subsequent agent behavior, which then tends to replicate itself.  Useful here are Maturana & Varela's notion of autopoiesis as well as Hermann Haken's concept of Enslaved States. Global order or patterns (that emerge through random behaviors conditioned by feedback) tend to stabilize and self-maintain.

Modeling Agents:

While the agents that inspired interest in complexity operate in the real world, scientists quickly realized that computers provided a perfect medium with which to explore the kind of agent behaviors we see operating. Computers are ideal for exploring agent behavior since many 'real world' agents obey very simple rules or behavioral protocols, and because the emergence of complexity occurs as a step by step (iterative) process.  At each time step each agent takes stock of its context, and adjusts its next action or movement based on feedback from its last move and from the last moves of its neighbors.

Computers are an ideal format to mimic these processes since, with code, it is straightforward to replicate a vast population of agents and run simulations that enable each individual agent to adjust its strategy at every time step. Investigations into such 'automata' informed the research of early computer scientists, including such luminaries as {{josh-epstein-and-rob-aztell}}, {{Von-Neumann}}, Stephen Wolfram, John Conway and others (for more on their contributions see also People on the upper right.

In the most basic versions of these automata, agents are considered as cells on an infinite grid, and cell behavior can be either 'on' or 'off' depending on a rule set that uses neighboring cell states as the input source.

Conway's Game of Life: A classic cellular automata

These early simulations employed Cellular Automata (CA), and later moved on to Agent-Based Models (ABM) which were able to create more heterogeneous collections of agents with more diverse rule sets. Both CA and ABM aimed to discover if patterns of global agent behaviors would emerge through interactions carried out over multiple iterations at the local level. These experiments successfully demonstrated how order does emerge through simple agent rules, and simulations have become, by far, the most common way of engaging with complexity sciences.

While these models can be quite dramatic, they are just one tool for exploring the field and should not be confused with the field itself. Models are very good at helping us understand certain aspects of complexity, but less effective in helping us operationalize complexity dynamics in real-world settings. Further, while CA and ABM demonstrate how emergent, complex features can arise from simple rules, the rule sets involved are established by the programmer and do not evolve within the program.

Agent Learning

A further exploration of agents in CAS incorporates the ways in which bottom-up agents might independently evolve rules in response to feedback. Here, agents test various Rules/Schemata over the course of multiple iterations. Through this trial and error process, involving Time/Iterations, they are able to assess their success through Feedback and retain useful patterns that increase Fitness. This is at the root of machine learning, with strategies such as genetic algorithms mimicking evolutionary trial and error in light of a given task.

competing agents are more fit as they walk faster!

John Holland describes how agents, each independently exploring suitable schema, actions, or rules, can be viewed as adopting General Darwinian processes involving Adaptive processes to carry out 'search' algorithms. In order for this search to proceed in a viable manner, agents need to possess what Ross Ashby dubs Requisite Variety: sufficient heterogeneity to test multiple scenarios or rule enactment strategies. Without this variety, little can occur.  It follows that, we should always examine the range of capacities agents have to respond to their context, and determine if that capacity is sufficient to deal with the flows and forces they are likely to encounter.

Further, we can speed up the discovery of 'fit' strategies if we have one of two things: more agents testing (parallel populations of agents) or more sequential iterations of tests. Finally, we benefit if improvements achieved by one agent can propagate (be reproduced), within the broader population of the general agents.