Group Smarts

Ideas gain currency, not because they're true, but because they're easy to recall and sound plausible at first hearing. Collective intelligence must be the best contemporary example I can cite without instantly being blacklisted by the CIA.

The idea of collective intelligence is based on the notion that simple organisms can (and often will) collectively exhibit behaviour that is much more complicated than what the individual organisms could muster on their own. Unicellular organisms band together and eventually turn into proper multicellular organisms. Higher order organisation of an astounding complexity is observed in ant colonies and bee hives. So far, so good?

Not really. The behaviour of insect colonies has been described in great detail, and holds comparatively little mystery when contrasted with the behaviour of a single ant or bee. Nobody has the foggiest idea of how bees with their one million or so neurons manage to navigate, memorise a path, express this as a dance and correctly interpret such dances by other bees. The behaviour of a colony is really simple and straightforward compared to that of a single individual. So-called emergent properties are emphatically not always more complicated than the properties of the constituent elements.

Although multicellular organisms have much more complicated behaviour than single cells, the extrapolation already breaks down at the next level with the simplest of animals, let alone humans. So what's going on here? Two observations should shed some light on this:

Humans spend most of their time not making conscious decisions. When we're walking down the street, we're not thinking about putting one foot before the other. There aren't really more or less intelligent ways of walking. We decide where to go and the rest of the process is goal-seeking and obstacle avoidance behaviour. Unless something interesting happens along the way we'll conveniently forget the whole trip. This is why crowd movements can be modelled using the same particle flow analysis methods as those used in fluid dynamics.

For all the complexity and speed of thought our brain permits, our output capabilities are extremely limited. We have a few hundred billion brain cells, and only about a hundred muscles under conscious control. At a higher level, many of the longest and most complicated deliberations result in simple decisions like what to buy or who to vote for. The same goes for more primitive organisms too, of course. The complexity of the behaviour of a collective of agents is limited, not by the individuals' internal complexity but by the bandwidth of their output channel. In terms of transmission rate, humans may have a significant edge over other animals in the form of speech and writing, but there too, the proportion of time actually spent talking about matters pertinent to the putative "super-organism" is extremely limited.

The upshot is that the collective intelligence of a random group of humans is not greater than that of a primitive organism consisting of the same number of cells.

We've not yet looked at what such a super-organism would look like, or what it would mean for us.

Here are some questions to ask oneself:

How many cells did we need before we became an intelligent species?
How many humans would it take to form a super-organism theoretically capable of thought or consciousness?
Are we conscious of our constituent cells? Do we care?
Would this consciousness have anything to do with the consciousness of its constituent humans? Would it care?
How can we be sure that just any superorganism is viable? Is the one we are part of one generation out of a successful line?

As humans we have one leg up over bacteria and ants. Humans are capable of abstract thought. Cells can't analyse their condition cellulaire. However, it does mean that we can't afford to lull our intelligence to sleep with the stupefyingly naïve idea that a self-organising collective of humans automatically figures out how best to take care of itself.

Monday 17 October 2011