Anticipation Without Prediction
This week and next the 16th edition of the Real World Risk Institute is being held (since the COVID event, on Zoom. Everyone of course misses gathering in person but we’ve managed to preserve the spirit, it seems.). I feel fortunate that I’ve been able to participate as an instructor over the past few years, and it is hard to believe it has been that long already — years. Life is long, yet time is short.
A topic that came up this week was the artificiality of what we call “prediction”. For a working definition: prediction is when we make some assertion about what we believe the future state of the system will be. Prediction is often treated as the only way in which we interface with the future, but this is simply wrong.
There is something perhaps not-so-obvious here in what appears to be an innocuous definition: the focus on the future state of the system.
In modeling systems, we establish a state space that describes the configuration of the system in the abstract at some moment in time. The classical case is the Newtonian state space: essentially, a list of the position and velocity of every ‘particle’ in the system at given moment. As the system evolves, the particles move, which amounts to the state space changing according to the ‘laws’ that generate the behavior of each particle.1
This idea of a state space naturally generalizes to describe a system in a way that is natural to an observer of the system: we might model the state of a clock by the position of its arms, for instance, rather than the trajectories of the atoms that comprise the material of the clock system. The ‘laws’ then of such a system could be driven by the way the gears and other mechanisms inside the clock drive the way the direction of the hands unfolds in time.
So, prediction is about specifying a belief that a particular state (or set of states) will be reached in the future by the system unfolding according to the ‘laws’ that drive it. And it follows that predictions are generally made by building a model of the system, running it forward faster than reality unfolds2, and seeing where the model goes before the system does.
But this is not the only way that agents interact with the future, nor is it the general way that they do — as I said above, it is in fact quite contrived and artificial.
Keep reading with a 7-day free trial
Subscribe to Applied Complexity Science to keep reading this post and get 7 days of free access to the full post archives.