The Ghost in the Ledger: On the Unseen Taxonomies of Public Data
When we champion the cause of open data, we often celebrate the raw datasets themselves—the spreadsheets of municipal budgets, the geospatial coordinates of park benches, the CSV files of public transit schedules. We imagine a world where this information, once locked in filing cabinets, is now free for anyone to download, analyze, and build upon. This is a noble and vital goal. But in our focus on the data’s liberation, we often overlook its silent, shaping companion: the taxonomy.
A taxonomy is the hidden architecture of information. It’s the system of categories, tags, and hierarchical relationships that makes data findable, sortable, and meaningful. It answers the question not just of *what* the data is, but *what kind of thing* the data is. And these systems are never neutral. They are human constructs, born from specific contexts, priorities, and biases. When a city government releases a dataset, the most profound statement isn’t always the numbers in the cells; it’s the labels on the columns.
Consider a simple, publicly available dataset on building permits. One municipality might categorize projects under "New Residential," "Commercial Alteration," and "Demolition." Another might use a more granular system, distinguishing between "Single-Family Dwelling," "Multi-Unit Affordable Housing," and "Historical Façade Repair." The raw data—the addresses, the dates, the costs—might be identical. But the stories each dataset tells, and the questions it allows a researcher to ask, are profoundly different. The taxonomy itself becomes a lens, focusing attention on certain civic priorities while leaving others blurred.
This unseen scaffolding presents a unique challenge for digital preservation and web archiving. We are adept at capturing the data file, the HTML page, the PDF report. But how do we capture the intent, the logic, and the institutional knowledge behind the taxonomy? A future historian looking at an archived dataset from 2024 might understand the values contained within it, but will they understand why those values were organized in that particular way? The taxonomy is the ghost in the machine, the context that gives the data its true significance, and it is often the first thing to be lost when systems change or staff turnover occurs.
Preserving open data, therefore, is about more than just ensuring the bits remain readable. It’s about preserving the meaning embedded within its structure. It requires archivists and publishers to document not only the data but the data’s data—the data dictionary, the classification rationale, the change logs for the categories themselves. Because the true record of our time isn't just the information we released; it's the often-unseen framework we built to make sense of it all.
Notes & further reading
A few pages I came back to while writing this: