Here is a leap (coming out of a discussion with my friend Dr. Slava Shumowski): one way to avoid both the “mathematisation” (the classical scientific approach), and the trivialisation (the “Realwissenschaft” [Carnap] approach – e.g. a set of case studies) could be to identify patterns – this approach has already been tried successfully in software engineering (“Gang of Four” – software design patterns) and in process management (process patterns, which have e.g. found their way into diagrammatic process description languages such as UML). This approach is also the basis of the recent (?) craze about Service-Oriented Architectures (SOA).
The search for patterns is a rather old human art form, of course – look for examples supporting this point.
“Search for patterns” might be a much more useful epistemological term than “scientist” or “search for knowledge” (or “truth” on that matter).