unable to be surprised ?

An architecture is, in some form, a model of the system that exists or will one day exist in real life. As stated by George Box (and many others): All models are wrong, some are useful.

In Bayesian probability, prior probabilities are used to

This means

Accounting for surprise means building architectures that integrate the prior probabilities into their model

Kaizen, architecture must either be flexible, or broad enough to account for adjustments in prior probabilities.