The posts can be read as being conditioned on the literal truth of Keynes and Whitehead’s view of the logic of uncertainty – empirical logic. But here I provide a more nuanced view.
By referring to Keynes, Whitehead, Good one:
- anchors the posts to a logically sound mathematical basis
- anchors to a body of theory that has been proven in large scale application
and hopefully (watch this blog):
- shows how a clear view about the fundamentals leads to a clearer view about derived concepts, e.g. in the foundations of science or ‘data fusion‘.
- provides some clear insights into the limitations of classical Bayesian probability, here considered as the special ‘single epoch’ case.
- support those areas of best practice in decision making that are ‘irrational’ from a Bayesian perspective.
But is this all there is to uncertainty? The classical literature deals with a single epoch. Using Whitehead et al we can deal with multiple epochs, and consider transitions between. But in the real world while we do seem to have punctuated equilibria, there can be unstable ‘wandering’ behaviour in between, for example during a crisis, during which we still need to operate effectively. Hence our ‘assumptions’ are ‘wrong’. So couldn’t we just stick with Bayes?
An informal justification is:
- The new framework is closer to the truth in terms of the uncertainty that it explicitly represents.
- In my experience it seems to fit better with how people actually make complex decisions.
- We could approximate to wondering behaviour by a succession of small epochs, about each of which one has little data. The very frequent uncertainty about the ‘jumps’ between epochs would then seem similar to having a continual change, if that should actually happens.
- An important class of contexts can be viewed as complex ‘games of games’, with Cybernetic co-adaptation. In this case one would expect real small ‘trial’ epochs.
- The theory shows that just because the world is not Bayesian doesn’t mean that one can’t usefully think logically, or that one needs to impose a solution.
Thus the Whitehead process model seems ‘close enough’ for us to be able to bridge the gap to any wandering between equilibria. Should we find a more appropriate model we could replace the Whitehead processes and work through the consequences. Similarly, the current model does not address the use of vague language, such as ‘short’.
Cynefin framework (maybe its ‘chaos’ corresponds to our ‘wandering’?).