Audit Approach

A spreadsheet was created to run an environmental scan of our digital collections. We created a profile about each collection and determined which corpus to pilot and which ones to prioritize in audit iterations. Given its diverse types of content formats, we decided to go with the FU Yearbooks Collection as our low-hanging fruit to assess.

FU Yearbooks Collection:

These collections consist of digitized copies of all the Furman University and Greenville Woman's College yearbooks, which include the Bonhomie yearbooks from 1901 to present, selected Entre Nous yearbooks from 1912-1931, and Blue and Gold (GWC) for 1901, 1902, and 1903. 

Audit Scope:

We established parameters to determine what constitutes auditable and non-auditable content. High-priority or auditable content includes:

Low-Priority or non-auditable content includes small student portraits, small faculty/staff portraits, indices, table of contents, blank pages (section dividers)

Overlaps:

Sections of collection were assigned at random, with some overlaps to maintain reliability and validity throughout the audit process. These random selections covered a wide range of time periods and content types (i.e., Bonhomie AND Entre Nous). Here are the overlaps needed:

• At least four raters per volume under review
• At least two areas of exploration per volume under review
• At least 20 pp of auditable content per volume (³10% of auditable content)

Sampling:

Although we did not up with a statistically valid sampling for our manual audit, we are confident we can draw valid extrapolations with a sample that represents at least 10% of the total collection. There are a total of 145 yearbook volumes, of which we should attempt to examine at least 15 volumes. See the breakdown below:

Part of this pilot phase was to figure out the minimum percentage of a yearbook volume to be reviewed. In other words, an acceptable indicator. During the last two rating sessions, we asked raters to glance through as many pages as possible for a limited period of time and to indicate the number of pages they actually audited. Determining a good sample depended on two factors (1) the volume size, and (2) the time allotted to read the pages assigned.

Potential Limitations:

Behaviors due to demand characteristics, that is, the cues that might give away the aim of this pilot study to participants, these could lead to raters changing their responses based on what they think we want to get from this audit. This in turn, can compromise validity and reliability. Replicability would likely be an issue, in that each digital collection is made of different content types, so the DEI audit survey may have to be modified, depending on the type of collection (Yearbooks vs Historical Images; Digital Collection vs Finding Aids) and the extent of the descriptive metadata (original content vs locally-supplied metadata).