What do people see in pictures? How do we describe images mashed up from stock photos taken out of their original contexts?

I thought of these questions as I read about the controversy surrounding the Dunkin Donut ad in which Rachel Ray was promoting coffee against a backdrop of cherry trees in bloom and a picture of a stately building. The black and white scarf around Rachel's neck conjured jihad-like associations in many conservative minds. Oregonians saw past the controversial scarf and asked what their state capitol building is doing in a Dunking Donut ad.

The Dunkin Donut ad controversy made me think of one of the more difficult questions asked during my interview for a metadata management librarian position last week: To what extent would I describe a digital object? For example, for the subject description, there could be many possible headings to describe any object but when is enough?

Traditionally, subject description has been limited to the content of the item - it's aboutness - and catalogers have some ways of weighing this aboutness which could be elusive as Patrick Wilson wrote in his essay Two Kinds of Power. Say, include a subject heading that at least represents 20% of the content of the book and cite the first heading that best describes what the book is about and which will be used to determine the classification for that book.

Now, with our emphasis on the needs of the users, we try to assign subject descriptors beyond content to the possible contexts in which the information resource will be used - what keywords would users likely to use to to be able to retrieve the resource? The possibilities are endless especially with images. The problem is partly solved by giving users both the options to search and/or browse. But to optimize search by including as many subject descriptors as applicable can use up very limited staff resources.

How do you delimit user needs and contexts? For an academic library, do you limit it to the information needs of students, faculty and staff of one department, do you expand it to the general knowledge domain under which that department belongs and hence consider users outside of that geographic locality, do you start thinking that the publication department might want to use that image for uses not related to the knowledge domain in which it was originally created for, and so on.

To what extent can cataloging/metadata practices control the ways by which people might retrieve or interpret information resources?