Emergence
Imagine a blank screen. As you watch, dots appear. The first dots seem randomly placed, but then you see clusters develop, mostly in the top and bottom halves of the screen. But right in the middle, running west to east, a series of dots turn up, sometimes alone, sometimes in small clusters of two or three. They connect into a meandering line. What is this? What explains the placement of the dots? Is there a force field? Are there attractors?
Then you are told the dots show the location of hotels, restaurants and pubs. The screen is a map. You immediately hypothesise that the clusters represent towns. But what about that meandering line? Slowly, the map fills with roads and towns. Rivers appear and for a moment you think the meandering line is a river — but then you realise it cannot be, as you see thin blue lines running away from the meandering line, ending up in fat ones at the bottom and top of the map. Then colouring and height lines appear and in a flash you understand there is a mountain range running west to east. The dots along the line represent restaurants sitting at the top of passes that allow roads to run north-south.
Anyone who has spent a holiday in the mountains develops a keen sense of where restaurants can be expected, and passes are top locations. Nothing like arriving at a pass after a strenuous drive, turning a thousand corners along a twisting road, to turn the mind to coffee!
Entrepreneurs seeking to open a restaurant are drawn to such locations as bees to a flower. A pass indeed is an attractor of sorts. It is as if the landscape exerts a force, a field that bends and shapes and forms hotspots in some places and is just about absent in others (on mountain tops, for example — unless a cable car runs to it).
Obviously, this force field is not a force of nature. It is something that emerges from the interaction of tourists with a mountainous environment with roads running through it. But the idea of a force field neatly captures our understanding of why there are restaurants in some places and not in others.
Much of our science follows a strategy of reduction. To understand a complex phenomenon, reduce it to its parts and try to understand each of them separately. This has worked very well and has allowed us to uncover the commonality in an enormous range of phenomena, ever widening as we dug deeper by finding smaller and smaller bits of matter.
But reducing is one thing, constructing is another. Intuitively, we expect that when you take something apart, it should be possible to put it back together. However, as most of us will have experienced at some point in their lives, this is not necessarily a practical possibility. Try deconstructing and reconstructing your laptop, for example. Yet we know it is possible — after all, someone made that laptop.
But try as you might, putting back together a dissected frog is a trick no one has ever pulled off (excepting drs Moreau and Frankenstein — in fiction). Is it possible, in principle? I tend to answer affirmatively. Scientists like George Church and Craig Venter have replaced significant parts of a bacterial genome with synthetic replacements, while these organisms kept on functioning. There seems to be no principal obstacle on the road towards synthesising an entire genome and putting it into an ‘empty’ bacterial shell.
This is not yet constructing per se, by a long way off. It is more like Theseus’ ship: replaced bit by bit over the years of his long journey, while retaining its function and identity.
Yet, constituting a living organism from inanimate parts makes us uneasy (which is the primary theme of the stories of Moreau and Frankenstein). After all, isn’t the quality of life itself of quite a different order than dead, albeit biochemical, molecules?
This is where the concept of emergence comes in. It gives a name to the observation that the whole can have qualities that none of the parts exhibit. The canonical example is a gas in a container. The gas has a temperature and a pressure, and both are related by a simple law to its volume (discovered by Boyle). We know that a gas is composed of myriads of molecules, each moving around in a chaotic tumble, coming into collision with the container’s walls. It is the sum of these innumerable collisions that exerts a measurable force on the container walls that we call the gas’s pressure. Moreover, the faster the molecules move, the higher will be the temperature shown on a thermometer inside the container.
The thing is, the gas has properties (pressure, temperature) that none of its constituent parts have. Molecules have neither temperature, nor pressure. To tie temperature to molecules, we say that the property of temperature emerges from the collective behaviour of the molecules.
We do not know how this emerging happens. We don’t have a clear understanding of the phenomenon: is temperature caused by the moving molecules, for example? Does one come before the other, in time? Do just two gas molecules exhibit pressure and temperature? If not, how many are needed? In fact, the use of the word ‘emergence’ is little more than hand-waving.
There are those that believe that intelligence is an emergent property, too. The reasoning goes a bit like this: clearly, humans are intelligent; clearly, this intelligence depends crucially on the brain and this we understand as a very complex composition of billions of similar parts, the neuron. So, analogously to temperature, intelligence may emerge as we lump together enough neurons. It is this reasoning that underlies the hope that artificial neural networks will, as the scale at which we can assemble them increases, exhibit true intelligence at some point —maybe consciousness, even.
Returning to the map of restaurants in mountainous terrain, we easily recognise the metaphorical use of the concept of ‘attraction of restaurants to mountain passes’. It is a convenient concept that aptly describes our experiences (as tourists) in the mountains. We do realise that other causes have been at work prior to the building of a new restaurant: entrepreneurial spirit, psychological attraction of people to high places in mountains, economic opportunity, practical considerations of cost and engineering. Yet, to capture all this we might still say that passes attract restaurants. We understand that to be a valid description on some level of what happens, even though more detailed explanations are possible — and are called for when we descend into specifics.
I believe that the concept of emergence should be understood as a relation between two different descriptions of the same phenomenon, where usually one of them is much more detailed than the other.
Warm and cold are primary sensations. This kind of ‘sensible temperature’ is available to us ‘natively’: it need not, indeed, cannot be explained to us in other terms. It is self-evident in the most literal sense of those words. Now, as scientists, we have built on that primary sensation by constructing thermometers. A thermometer is based on the observation that the length of most materials increases as their temperature increases. This we can establish first hand, by changing the temperature of a material in a controlled fashion and measuring its length. In the limited temperature range that we can experience directly it is only under special circumstances that we can actually see its length change. But it can be done and by extrapolation we’ve discovered how extremely high or low temperature may rise in some materials.
A thermometer enables us to use our eyesight to gauge temperature of objects far outside the range of our direct experience of heat and cold. In a similar fashion, we’ve designed pressure gauges. Using these instruments and others, over the course of centuries of measurements and experiments we’ve come to an understanding of a gas in the terms I’ve used above: tiny, invisible molecules flying around.
This description of a gas is very different from the one that our senses offer us: that of a cold or warm wind, for example. Yet, we know both are appropriate descriptions of the same phenomenon. And then we are confronted with the conundrum of the emergence of temperature and pressure — our immediate sensations — from molecular movement and collisions.
Why is it a conundrum? I think it is because we have attached, due to its success, more importance to the physical theory than to our immediate sensory experience. After all, we know how limited our senses are and how easily they are fooled (just think of visual illusions). To the scientific mind, molecules and atoms are more real than tables and chairs.
But in the scientific account of the temperature of a gas, sensible temperature has been carefully edited out. Remember the thermometer? It enables us to do away with accounts based on our sense of temperature and to replace it by accounts based on eyesight — aided by the thermometer, of course.
So we’ve maneuvered ourselves in the position where we rely on accounts of the world around us that are based on eyesight, aided by instruments — and then we wonder how our sense of warm and cold fits in! Somehow, we feel ourselves obliged to think those secondary properties have to arise out of the account that we deem primary.
But it cannot be done! The molecular account is, by construction, devoid of concepts based on those senses. Then, all that rests us is to declare that temperature emerges from molecular movement. Hand waving, really: nothing more.
The concept of a restaurant attractor is playful, it does not worry us. We don’t take it too seriously. We let it arise, use it when appropriate, trusting that we’ll stay on the safe side of the line between mental health and delusion. No one will ever feel they are forced against their will to build a restaurant on a mountain pass because of this mysterious attractor.
But it’s a quite different story with the emergence of temperature, let alone of intelligence. Fierce battles are fought on these grounds. Why is this? Because, on the one hand, we’ve come to see physical theory not as just that, a theory, but as the truth. This is what reality is: a composition of quarks (or strings, if you’re inclined that way). But, on the other hand, even the most reductionist mind will recognise that she is, well, just that too: a mind (though some people go so far as to call that an illusion— I will call that a mental problem!). Mental phenomena undeniable cause things. But as cause and effect are understood as the prerogative of the physical, we think mental phenomena have to arise from the physical.
As we don’t have a clue about how that could be — and it cannot be, because we’ve edited out, very carefully, the mental from our physical theories — we fall back on emergence by way of explanation. But, again, it’s a weak one, not more than hand waving!
All this disappears if we accept physics as a description of the world and not as the way it actually is. Then, a report of our first hand experience of the mental is a valid description of our experience — and a physical explanation is another. When both kinds of descriptions apply to the same phenomenon, there may be occasions that saying that a concept in one descriptive domain emerges from concepts in the other descriptive domain actually elucidates the relation between the two. But there is no a priori reason why that should be the case.
Another perfect example, that has lain hidden in my memory for many years, is due to my father. A professor emeritus in biology, he dedicated his research career to vertical migration of Daphnia species. You’ll be forgiven for not knowing these animals to be water fleas, tiny planktonic crustaceans. Daphnia feed on algae, that in turn need light to photosynthesise and grow. As sunlight is increasingly blocked as it has to pass through more water layers, most algae grow in the upper layers and consequently this is where Daphnia grazes. However, in early spring, fish eggs hatch and the young fish prey on Daphnia. From then on, the water fleas start to migrate vertically; that is, the entire population rises to the surface at night (when the fish cannot see them) and dives to deeper layers at day. Clearly, migration of the Daphnia population is caused by the fish.
Now, you’ll agree with me that a Daphnia population consists of Daphnia individuals. This implies that migrating populations equals migrating individuals. And, taking it one step further, you’ll conclude that changed population behaviour implies changed individual behaviour.
But this is where you would be totally wrong. A Daphnia that is close enough to perceive a fish, is lost. Consequently, a Daphnia individual cannot learn that the fish population increases and adapt its feeding strategy accordingly. Daphnias don’t know there are fish.
Instead, what happens is that the fish selectively remove the individuals that do not migrate. Those that do (and there is always a tiny minority that does), find their competitors removed, feed more and flourish. They come to dominate the population. And this causes the behaviour at the population level to change.
A growing fish population causes a growing Daphnia population to change its behaviour; but this is in no way attributable to changing individual behaviour. A cause at population level is not a cause at individual level. *)
Perspectives is not based on reductionist thinking. It does not rely on an account of semantics in terms of reference to some ‘elementary’ things and their composition. It is a descriptive language whose justification lies in its pragmatics: one can do useful things with it (build information infrastructure support for people trying to co-create a process while being distant from each other).
The Perspectives modeller can freely design roles and properties in contexts to his liking without having to bother about how it relates to other descriptions of reality, as long as it fulfils the purpose stated above. This depends crucially on the understanding the user roles have of their context — and user roles are modelled, too.
This is not just a quirk. There could not be a greater contrast to the (usually implicit) notion of semantics underlying contemporary information systems. These are built on information models that claim to be a description of the world as it is — rather than the world as how it is understood by the participants in the process.
In other words, contemporary information systems claim more or less universal truths about the world, while Perspective models are far more pragmatic: they are ‘valid’ insofar as they work for their users. The users, that is, in a small context. This is crucial: we may model the same part of the world differently for other users in another context.
Missing that limited context, information systems invariably run into problems as they come to cover larger and larger parts of the world (think enterprise systems or logistic chains, for example) — simply because people that do not work together need not conceptualise the world together.
- ) A personal footnote. In his later years, my father wrestled with the mind-body problem. As a biologist, he tried to show that mind must arise from biological processes — arise as in ‘emerges’. I find it sweetly ironic that he provided me with the perfect counter example of the very concept of emergence he invoked. He sharply distinguished, as he called it, proximal and distal causation. A distinction that he taught tirelessly to many generations of students.
This is the fifteenth column in a series. The previous one was: Guns, Germs and Databases. Here is the series introduction.