Back reading about ASK....
The more things change, the more they stay the same. I've probably read this a few times before but it's always worth a re-read:Belkin, N. J. (1980). Anomalous states of knowledge as a basis for information retrieval. Canadian Journal of Information Science, 5, 133-143.
Belkin's thing in this time period was the ASK model - anomalous states of knowledge. It's anomalous because of "inadequacies in a state of knowledge can be of many sorts, such as gaps or lacks, uncertainty or incoherence, whose only trait is 'wrongness' " (Belkin, 1980, p.137). In other words the user may not know exactly what they need or what will solve the problem. They may not be able to specify the need at all (cognitive specifiability), be able to put the need in terms the system will understand or the system might have a poor representation of the text (linguistic specifiability). So here's the kicker -- IR systems (still!) are frequently designed based on what the text representation looks like, not based on what shape the users are in, how well they know what they need and can convey what they need.
So we're still looking at pure topicality and text matching for relevance, in some cases, and testing the system with known-item searches. Every system by now should easily be able to retrieve known items.
This is why librarians get excited with faceted presentations like in the TLC, Endeca, EI, and Ebscohost products (turned on late this week, not sure if Ebscohost upgrades are as good as I was hoping but anyway...). It's *not* because we just want more toys (although we do) -- it's because ASK-type situations are more prevalent in our reference transactions than known item searches (once you've taken directional and fix-the-copier questions out of the stats). We're so like goalies today: we don't get the easily findable, we get it after they've tried what they know or when they're in a new area.