Primary Sources in a Deepfake Era

Using first-hand sources to get at what's real

This is the ninth in the series Get Your Facts Straight: Research Skills for Writers. For more about this 18-part series, including the complete schedule and the Table of Contents with links to all the other articles in the series, click here.

---

You may have encountered the term “unreliable narrator.” It’s a term often used in fiction to refer to those narrators whose version of events, within a story, might not be completely reliable. You might’ve encountered an unreliable narrator in a mystery or a spy novel, putting just enough of a spin on things to be a bit (or a lot) confusing. The unreliable narrator’s job is to heighten the tension in a story, making things just weird or unclear enough, or throwing in enough alternative possibilities and red herrings, to raise a bit of frustration and sharpen the reader’s desire to get to the bottom of things.

Real life unreliable narrators are a lot less fun. In fiction, the author can simply clarify the information, debunk the lies or rumors or half-truths, and tie things up in a bow for the reader or viewer. When we’re dealing with unreliable sources of information in the real world it’s a lot harder—no deus ex machina is going to help you—and the stakes can be much, much higher. Figuring out what’s true versus what’s only got a kernel of truth versus what just sounds plausible is essential. It’s also difficult and time consuming.

Not all unreliable narrators intend to be unreliable. Anyone who’s ever played a game of “Telephone,” heard an urban legend, or watched a meme morph as it circulates around social media is familiar with the ways people can innocently introduce wrong or misleading information into an anecdote or story.

Norman Rockwell, “The Gossips,” 1948

But sometimes unreliability is not innocent at all. Misinformation, disinformation, propaganda, “spin,” and outright lies told in the interest of any ulterior motive don’t just hurt people and harm societies, they damage our ability to place our trust in learning, information, and meaning. It is a moral injury.

At this moment in time we live in an information ecosystem where it is exceptionally easy to generate and disseminate misleading and false information. We also live in an information ecosystem where it has become more and more difficult to tell what’s real and what isn’t, what’s true and what’s a lie, and where systems like AI, which cannot evaluate the validity of information, compound misinformation and lies through pure mechanical, uncritical repetition.

When it comes to our ability to do research and locate trustworthy, valid information, this can be devastating. What happens to our ability to find and share solid information, especially first-hand and eyewitness information, in an era where deepfakes—faked videos in which individuals are made to appear to say and do things they never did—are becoming more common, as are other forms of egregious forgery and fakery? More to the point, what can a researcher do about it?

Primary Sources and Technological Shifts

One thing researchers can and should do is to think carefully about primary sources and learn to be canny about selecting and using them.

A primary source is any first-hand contemporaneous artifact or documentary source of an event or topic, especially those produced by people who were directly involved in or witness to the events in question. Documents like diaries, letters, newspaper articles, original artworks, eyewitness accounts, sets of statistical data, audio or video recordings, and so on provide original sources of information about a particular event or moment in time.

Once upon a time, primary sources were relatively stable once created. This had to do in large part with the technologies used to create them. Once you wrote that letter or that page in your diary, it wasn’t actually all that easy to change what you’d written — you’d have to destroy the original and write another. The same was true of things like newspaper articles or audio recordings made with physical media (vinyl records, for instance, or magnetic audiotape).

Once the object was made, whether that meant printing the newspaper or making a copy of an 8-track tape, altering it was not impossible but was unlikely, simply because the alterations were physical and left distinctive traces. Changes to printed materials and physical recordings leave funny-looking visible spots in printed materials, audible glitches and skips in recordings, visible splices in tapes. For similarly pragmatic reasons, documents for which there are many physical copies—printed newspapers or books, physical photographs or audiovisual recordings—are less likely to be doctored after they are produced because of the difficulty of doing so to all the copies.

Audiotapes were mostly spliced for legit reasons. It’s how audio recordings were edited, once upon a time.

Every once in a while, a really gifted forger or counterfeiter manages to create a undetectable (or nearly undetectable) physical fake, but that’s always been pretty rare. It’s difficult and often expensive to produce really good fakes of physical objects, so that sort of thing has most often been found in association with objects of particularly high economic value: counterfeit currency, fake jewels, art forgeries.

An important principle in assessing the possibility that a source or document is fraudulent or even just misleading is quite simply to consider whether passing it off as genuine would be valuable. If a fraud wouldn’t pay off in some way or other, it’s less likely that anyone will go to the trouble. On the other hand, when fraud becomes easier to perpetrate, the cost/benefit ratio of going to the trouble to manipulate or falsify things changes.