Dictionaries are a great example. They are a useful general-purpose abstraction. Often times programmers will develop a hierarchy of classes that once you get to the root of is just a dictionary with some window dressing to hide the fact that it's just a dictionary. Sometimes it's easier to expose the dictionary directly and code the special cases directly. The programmer may fret about tight coupling but if you know you're never going to replace the dictionary-based implementation, it can be a lot easier to understand, especially if it's one-off code. And when dictionaries are provided by your language/libraries, you don't have to unit test them at all; you assume they work.
But now your program is weakly typed. Using the same logic, maybe your id is just an integer, but having an "ID" type will prevent bugs in the future, just like "dressing up" the dictionary with a meaningful name and exposing only the minimum amount of methods necessary will avoid people being tempted to, say, add things they shouldn't in your dictionary, not to mention that strong typing has a documentary value.
