The Archive

Show notes

To establish what happened in the past, a historian almost never has the event — they have a record of it, and that record has survived a long, untidy journey to reach them. The archival professions exist to manage that journey. Their first question is rarely is this document accurate? but where did this come from, who held it, and how do we know it is what it claims to be? Provenance — the documented origin and ordered chain of custody of a record — and authentication — the demonstration that a thing is genuinely what it purports to be — are not bureaucratic flourishes. They are the conditions under which a future reader is entitled to treat a record as evidence at all. A document and a reliable record are not the same object: the document is the artifact, while the record is the document together with the context and custody that make it trustworthy.

What the discipline has learned, over a long history of working with fragile and contested material, is that the hard part is preserving the conditions rather than merely the bits. Context decays faster than text. A letter without a date, a sender, a file it belonged to, or knowledge of how it entered the collection can be perfectly legible and almost worthless as proof. So archivists invest in description, in maintaining original order, in recording who accessioned what and when, in keeping the surrounding material that lets a later reader reconstruct meaning and gauge reliability. The aim is to leave behind not just the words but the apparatus a careful, skeptical reader of the future would need to reconstruct the record and decide how far to trust it.

Set that beside the records machines now produce. They are voluminous past any precedent — logs, traces, outputs, intermediate states — and yet they fail the archival tests almost completely. They are typically un-authenticatable: nothing binds an output to the exact model, data, and conditions that produced it. They are mutable: rewritten, regenerated, or silently overwritten with no surviving prior state. And they are contextless: a string of tokens detached from the provenance that would let anyone establish what it is or whether to believe it. The danger is not too little data but a flood of data that can never become a record — abundant, plausible, and inadmissible the moment someone asks where it came from and whether it has been altered since.

That is exactly the safeguard this show is about. A conclusion is legitimate not because it was reached but because the path to it can be reconstructed and contested later, by someone who was not in the room. Provenance and the reconstructable record are the whole point — and the archive is where the human record of how to keep them has been worked out across centuries. The book this series accompanies, Admissible Reality, argues that machine-made evidence has to meet a comparable bar: not just an answer, but a record whose origin and integrity a future reader could establish. The episode asks an archivist what their craft already knows about building records that survive — and how little of that knowledge the machine-made trail currently embodies.