Records are made by people
A digital exhibit usually draws on records that were written long before the exhibit was imagined: accession registers, catalog cards, donor files, old labels. These records feel factual, and a database field with a neat heading looks even more neutral. But every record was written by a person at a particular time, with particular assumptions about what mattered and who was worth naming.
When those records move onto a screen, their slant moves with them. An exhibit does not have to say anything biased to repeat a bias; it only has to copy a description without questioning it.
Two kinds of problem: gaps and slant
It helps to separate two issues, because they call for different fixes.
Gaps are about what is absent. A collection assembled by wealthy collectors may hold fine objects from the powerful and little from ordinary households. A museum’s holdings may reflect what earlier generations chose to collect, which often means certain regions, genders, classes or communities are thinly represented. Photographs often show those who could afford a portrait or were considered important enough to document.
Slant is about how what is present is described. Examples include:
- Outdated or loaded terms used for groups of people, which were once routine and are now considered offensive or inaccurate.
- One-sided descriptions, such as an object described only from the viewpoint of the person who acquired it, not the people who made or used it.
- Missing names. A photograph of a worker captioned only by the employer’s name, or a craft object attributed to a “native artist” rather than a named maker or community.
- Assumptions about importance, where the activities of some people are described in detail and others in a line.
- Gendered or cultural framing, treating one group’s practices as the default and another’s as exotic.
Where bias enters the pipeline
Bias is not only in the original record. It can enter at several later steps.
- Cataloging. Subject headings and descriptive standards reflect the period they were written. Some are updated slowly.
- Digitization choices. If the most photogenic or best-known objects are scanned first, the online collection appears even more skewed than the physical one.
- Selection for the exhibit. Curators choose what is displayed, which can compound earlier gaps.
- Search and tagging. Automated tagging or sorting built on historical text can repeat its patterns. A search that returns mostly one kind of person tells visitors something, whether you meant it or not.
- Writing the label. Shortening a complex record for a screen can drop the qualifications that kept it accurate.
Good collection metadata practice includes being clear about whose voice a description reflects and when it was written.
Questions to ask of a record
Before a record becomes a label, a few plain questions go a long way:
- Who wrote this, and when? Does the wording reflect their point of view?
- Whose name or voice is missing, and could it be added?
- Is any term here one that the community described would not use today?
- Does the record state facts, or opinions presented as facts?
- Who benefited from the object being collected, and under what circumstances? Was it a gift, a purchase, or something taken?
- What does this collection not include, and is that worth saying?
These are not about proving that previous generations were wrong. They are about not repeating a view as if it were settled.
Practical ways to respond
Keep the original, add context. Quoted historical wording can be shown, clearly marked as historical, with a contemporary note explaining why it is not used now. Quietly rewriting records erases the evidence that the slant existed. On the other hand, repeating an offensive term in large type as a headline helps nobody. The balance usually lies in marking, framing and sometimes warning, which links to the broader practice of handling sensitive and distressing content.
Name people and makers. Where records allow, replace anonymous descriptions with names, and where names are unknown, say so honestly rather than inventing a generic label.
Ask the people concerned. The most reliable correction for a description of a community is that community’s own voice. Consulting source communities can supply terms, context and stories that no archive holds. Treat such input as a contribution that deserves acknowledgment, not an unpaid proofreading task.
Say what the collection leaves out. A short note such as “this collection mostly documents the town’s merchants” is more honest than a title implying it covers the whole town. Admitting the limit makes the exhibit more credible.
Use varied sources. Oral accounts, community photographs and other kinds of evidence can balance documents that were created by one side. Recording oral histories is one way to bring in voices that official records omit.
Review regularly. Language changes. Build in a point at which labels and records are looked at again, and a simple way for visitors and colleagues to suggest corrections.
Be careful with automated tools
Automated description and tagging tools are increasingly offered for large collections. They can save time, but they learn from existing text and images, so they may reproduce the same slant at greater scale, or invent confident descriptions where the record is silent. If you use them, treat the output as a draft that a person reviews, and label it honestly where it appears.
Accept that perfection is not available
No collection is neutral, and no label is final. The goal is not an unbiased exhibit, which does not exist, but a transparent one: it is clear about its sources, honest about what it lacks, and ready to change when it learns something better. That attitude also tends to build trust with visitors, who are generally more comfortable with “we know this is incomplete” than with an unexamined claim of authority.