David Medeiros and I got into a discussion in the complex printing thread that I thought deserved to be a new topic:
David Medeiros, in various posts that I've collapsed below, said:
Too much detail, even when it is accurate, can create a map that is less effective than one that is generalized to suit its purpose.
Krygier and Wood talk about it a bit in Making Maps (Map Generalization and Classification). They discuss the idea that maps work by "strategically reducing detail and grouping phenomena". It's the reducing detail we're mostly about here.
It is, to me, a bedrock concept within cartography but one that most GIS map makers have a very hard time with, especially in science based GIS work. The assumptions in GIS are a) GIS and GIS data are inherently very accurate, and that simplifying the display of your data is somehow dishonest.
Maps intended for sharing and communication end up with a mess of raw data on them that clouds the actual message being communicated and in the end there is a potentially misleading message of accuracy that may get transmitted when in reality GIS data can often be much less accurate then manual mapping, just a lot more precise.
For me the use of generalization hinges on the concept of the maps communication goal. If transmitting raw data is the goal then obviously don't generalize, but that's a poor use of a map in most cases as only the analyst really needs to have all the data at hand for their work. Instead the map should be communicating your results or findings and for that it's probably better to simplify.
I think nearly everyone would agree on the kind of smoothing that deletes outliers for boundaries such as animal ranges or forest cover, because those make the map more understandable at a glance.
Where I'm having trouble is justifying simplifying something like a municipal boundary or national forest polygon from 500 vertices to 50 when 500 is overly complex for the scale. In my gut, I absolutely believe it's worth the effort and makes the map better, but I fear that it's really an aesthetic preference based in old production methods. Same with watercourses: is there really any psychographic evidence that readers have an easier time following my smooth, sinous uniformly 4 pt Missouri River instead of one that USGS compiled at 1:10,000 that shows every sandbar and inlet?