"Real Design" Problems

Characterized by the subtle, qualitative interaction of an uncountable number of effects, integrated with a very small subset of quantitative, "designable" variables.
Controllable and measurable variables are usually misleading if used to manipulate and evaluate the overall quality of a system.
The measure of success is more a "feeling," how much people "like" what you have done, than anything else - and that in itself is very time and context-dependent.
The only viable approach is to use a depictive language (build scenarios, build realistic models, build prototypes) and to "live" with it - to experience it directly in it's real-world context.