There has been a great deal of attention paid to data cubes, cube computations, and the like, in the context of OLAP. It is generally recognized that most OLAP users study the results of cube computations as a first step in determining where drill down is required. However, the assumption I made (and I believe most people in the database community) is that this drill down process is an inherently human endeavor, and there is little the database could do to help (beyond speeding up computation).
This paper is interesting because it attempts to have the system "drill down" on behalf of the user and provide the meaningful explanations a user may be seeking. I do not have a clue whether the proposed system gets it "right". Also, this paper follows the general theme initiated by the author, with Agrawal and Megiddo, in EDBT 1998 . Yet, this paper is worth a read because it makes a credible attempt at expanding the boundaries of OLAP systems.