The Ghost in the Machine: On the Erasure of Context in Public Records

We celebrate the digitization of public records as an unalloyed good. The narrative is familiar and comforting: dusty boxes in forgotten basements are transformed into gleaming, searchable databases. Accessibility is increased, transparency is served, and the public good is advanced. This is the received wisdom, and on its face, it is difficult to argue with. But in our rush to applaud the act of digitization, we often overlook a more insidious, quieter loss: the systematic erasure of context.

Consider a single paper memo from a municipal planning department, circa 1985. As a physical artifact, it exists within a rich ecosystem of clues. The weight and texture of the paper, the fading ink from a specific brand of dot-matrix printer, the coffee stain in the corner, the handwritten marginalia from a colleague—all these elements are non-content data that speak volumes. They tell a story about process, about urgency, about informal networks of communication that operated outside official channels. They are the ghost of the organizational machine that produced them.

Now consider its digitized counterpart. Through a high-quality scanning process, its text is perfectly captured, its metadata logged. It is now a PDF in a vast digital repository, instantly retrievable by a keyword search. And yet, it has been flattened. The stain, the handwriting, the physical wear—the very things that made it a unique historical witness—are deemed non-essential and discarded in the transition. We have preserved the information but amputated its soul. The record is now a sterile data point, stripped of the humanity that created it.

This is not a call to halt digitization, but a plea for a more nuanced understanding of what we are doing. The prevailing wisdom treats digitization as a simple act of translation from one medium to a better one. In reality, it is a act of curation, of interpretation, and yes, of loss. We are making conscious decisions about what aspects of a record are "worth" preserving, often defaulting to a narrow definition of content that privileges machine-readability over human meaning.

The danger is that we are building a future historical record that is paradoxically both vast and shallow. Researchers of the future will have access to every word ever typed by a public official, but they may struggle to understand the culture that produced those words. They will see the decision, but not the debate scribbled in the margins; they will see the policy, but not the passion that stained the paper. In our quest for perfect, clean data, we risk creating a digital ghost town—a perfectly preserved set of facts waiting for a context that we systematically deleted. True preservation requires us to look beyond the text and find ways to capture the ghost in the machine.

Notes & further reading

A few pages I came back to while writing this: