The Idea of Semi-Permeable Agents
By Greg Fisher
A couple of months ago I attended a fascinating workshop organised by my colleagues David Hales, Jeff Johnson and Jerey Pitt, on the subject of ‘agents’ and ‘agency’ within the context of complex systems and computer simulations. The discussion was excellent in part because of its content but also because of the people I met and the work they’re doing.
During the discussions, something kept jarring within me – a quiet voice was telling me there was something not quite right with some of the language being used. I realised it was in the domain of what my philosopher colleague, Rhett Gayle, refers to as atomism, which is when agents in some system are viewed as fixed, with an impermeable boundary surrounding them (in effect or in reality). The term comes from the very old idea that atoms were the smallest ‘things’ in the universe, making them indivisible by definition. We now have quantum theory, which tells us that atoms aren’t in fact the smallest ‘stuff’, however the concept of atomism has remained.
An important critique of Complexity theory (or, rather, how it is being practiced by many) is that some complexity theorists are being atomistic about how they think about people. It was this that was gnawing at me at the workshop. I thought I’d articulate these thoughts here and also introduce the idea of the semi-permeable agent, which I believe would help us move away from atomism when thinking about people in complex social systems.
We can start by looking at John Holland’s emphasis on semi-permeable boundaries in complex systems (see, for example, his book “Signals and Boundaries: Building Blocks for Complex Adaptive Systems”). The idea is self-evident: some boundaries are impermeable, which means nothing can penetrate them, neither escaping from within nor entering from outside. And, at the other extreme, we can think of a perfectly permeable boundary where everything penetrates. Of course in extremis this means there would be no boundary at all but it’s a useful idea nonetheless.
In between we are in the world of semi-permeable boundaries, which let some things ‘in’ – but not all – and they let some things ‘out’ – but not all – allowing whatever is inside the boundary to evolve internally. Here, ‘things’ which might transfer across the boundary might include matter, energy, and information. Clearly, if there is a two-way flow across the boundary then we can see very easily how the semi-permeable agent and everything it interacts with will co-evolve.
The traditional approach to agency within social sciences, including economics, is to treat human agents as if they had impermeable boundaries, which is an atomistic approach. For example, standard microeconomics treats agents as if they had fixed preferences and a given budget. They are not influenced by other agents.
Cognitively and psychologically we ourselves make similar simplifications in our every day lives. The abstract word ‘person’ is a simplification which engenders the idea that we are discrete entities. And we are given names to distinguish us from other people, implicitly drawing a boundary around each of us.
So, in the West at least, we seem biased toward thinking of people (and other animals and objects) as impermeable, bounded entities. We can add ‘context’ around some entity but more often than not we think of the entity as if there were a boundary between it and its context.
This form of abstraction has a lot of use – it has helped our ancestors make sufficient sense of their world, allowing ‘them’ to survive and reproduce. And this process of abstraction isn’t wholly “wrong”, although I would point to it as a form of ‘coarse-grained cognition’, a term coined by Murray Gell-Mann. When we look at human agents more closely – more fine-grained if you like – we see that reality is more, well, complex.
Of course, at the other end of the spectrum, if humans were perfectly permeable then the notion of a human would be irrelevant and we wouldn’t have a term for it. My colleague David Hales wouldn’t exist, for example (which some people might be ok with). Of course, in our everyday world, the idea that we don’t really exist would be absurd.
It is preferable, I would argue, to think of human agents as semi-permeable. On the one hand, imperfect permeability leads us to recognise individuals existing in their own right; but they are not impermeable as if atomism were true of people. We are somewhere in between: we are semi-permeable.
We can unpack this some more. It should be obvious that we exchange matter and energy with the rest of our environment through consumption, waste, breathing, etc. But there is also an exchange of information – we sense things and we also give out signals, allowing others to sense what we’re doing. Of course, this is true of all organisms, not only humans.
Complex Social Systems
Delving further in to the realm of people and information, it is clear that we use this sensed information to build quite sophisticated internal models of the ‘outer’ world. Our central nervous systems have evolved to recognise patterns and to deduce things, and our imagination can anticipate potential futures and alternative circumstances. With this in mind, and thinking about groups of semi-permeable people, we can see that the norms, values and moral principles we hold are best thought of as socially constructed by semi-permeable agents.
For example, and on the whole, people in British society value freedom of speech (slander being a typical exception) – it’s a value many of us absorbed from others when we grew up and it is something we most of us perpetuate (others absorb it from us). How we frame things – which we might think of as our cognitive architecture – is also co-constructed by socially emergent processes. This is not just about what we might learn at school or university, it is inherent in the every day language we use.
The processes involved in our internal modelling are both fascinating and complex. Our sub-conscious seems to play a significant role in our values framework, which has a substantial influence on how we ultimately behave. We might think we make reasoned, dispassionate choices with our conscious brains all of the time but our subconscious brains have a substantial role to play in the framing of, and values implicit within, our conscious thoughts. And this sub-conscious architecture is substantially – if not entirely – socially constructed. This is a very different view of agency to that we see in economics, for example.
To conclude, complexity theorists have inherited the idea of impermeable agents from a long history of reductionist science. History matters, and it is always tempting to revert to what we know. To move in a better and new direction, I think we need to think of (and model) human agents as semi-permeable – I offer this as a better (though still coarse-grained) approximation of human agency in complex social systems.