One of the attributes that many (if not all) complex systems have is the ability to change (customise) them in controlled ways. Indeed this, by definition, is why some systems are complex (that is exhibit  emergent behaviour) because sometimes the users of those systems select options or make choices that enable unexpected behaviour to occur. Giving users such choices clearly has a number of implications on the architecture of a system.

1. Users have to make a choice, no choice is actually a choice itself as it means they are choosing the default.
2. Making systems customisable assumes users have the will (and the time) to decide which options they want to change. Many times they don’t so the default becomes very important as it will dictate how the users actually use the system, possibly forever.
3. The more options that are built into a system the more difficult it becomes to test each potential combination of those options. Indeed there comes a point at which it becomes impossible to test every combination (at least in a reasonable amount of time), hence the importance of beta software (let the users do the testing).

In his blog entry Triumph of the Default Kevin Kelly points out how “the influence of a default is so powerful that one single default can act as a very tiny nudge that can sway extremely large and complex networks”. The oft quoted example of how defaults can influence behaviour is that of organ donation. If you make the donation of organs upon death automatically an “opt out” choice (it happens unless you refuse beforehand) versus “opt in” (it does not happen unless you sign up). A opt out donor system greatly increases the number of organs donated.

For complex systems then, the default architecture of the system becomes very important. The choices the architect makes on behalf of the users of that system will not only dictate how the users will actually use the system but may also influence their behaviour (in both positive and negative ways). Defining the defaults is as important an architectural decision as to what technologies should be used and sufficient time should always be planned in to allow such decisions to be made in a sensible way. The system architecture that results can profoundly affect how that system will be used.