Recommendation systems may be used to help predict preferences and behaviors of individuals. For example, recommendation systems may help select items that are of interest to the individuals. The recommendations may be based on items that are often selected together by other users of the system, which may increase the relevance of the recommendations to the individual users. In many instances, the recommendations may be organized in a hierarchy of categories or types. One of the problems with current recommendation systems is that they often provide recommendations that are not tailored to the needs of the individuals. For example, an individual may not need or desire items of a particular type or category of the hierarchy, and consequently these items may not be included in the user's recommendations. This results in lost opportunities for the user to discover items they may have been interested in or may provide recommendations that are not relevant to the needs of the individual. The approaches described in this section are approaches that could be pursued, but not necessarily approaches that have been previously conceived or pursued. Therefore, unless otherwise indicated, it should not be assumed that any of the approaches described in this section qualify as prior art merely by virtue of their inclusion in this section.