Monday, July 31, 2017

FSA/OWI Archive, The Hole-Punch, and Photoshop

Those of us who are more or less familiar with the history of the FSA/OWI know that Roy Stryker used a hole punch to "cancel" negatives that he deemed not suitable for printing. So, there's a big pile of negatives out there with little holes in them. This is often reported as a Historical Travesty etc, and there are sometimes Dark Hints that Stryker cancelled negatives that did not support his propagandist program.

All of this is absurd. Most of punched negatives I have seen have the hole jammed through somewhere obvious but otherwise at random. Stryker wasn't trying to destroy information, particularly. Indeed, the negatives remain archive. To properly destroy the information, Stryker would have shredded them or similar, and he certainly wouldn't have carefully preserved them and arranged for their transfer to the Library of Congress. Techniques for destroying film certainly existed and were used by the government.

What Stryker was doing, at least mostly, was picking out the good ones. The propaganda campaign, which was very real, was executed largely by giving his team shooting scripts.

In some cases, a face is obscured, perhaps a sign or similar important detail is chopped out. In many cases, of course, another negative from the same group will be available, un-punched or punched in a different way. The point is that, in general, the punched hole doesn't really delete anything of historical interest.

Which leads us to Photoshop. This chappie at the University of Connecticut said to himself "what would happen if we used Photoshop's content aware fill to fill that hole back in?" and the answer is, obviously, "it does OK as long as there's nothing interesting in the hole". See his article here.

He goes on to speculate about what this might all mean. You'd think that as the Head of Digital Imaging and Conservation at a University he'd think this stuff through a bit, but no, what we get is this:

Yet, what are we to make of these surrogate negatives? Though they are not based upon standard content replication like their hole-punched brethren, the software-filled versions still hold a certain spell and feel of visual symmetry. From an archival standpoint we may even regard them as born-digital objects in their own right. Perhaps in the end they may simply be best considered informed guesses on fragments of displaced history.

"born digital objects in their own right," how terribly cute! What they are, sir, is they are bloody dangerous and little else. The trouble with these damned things is that part of them is false. Part of them is not indexical, it is guessed at by the computer, and we don't know which part. The thing looks and smells like a photo, and most of it is a photo, so we trust it. Part of it, an unknown part, isn't a photo. More or less by definition this process cannot add to the truth value of the photograph, it can only subtract. Taking historical artifacts and making them less true is probably not a great idea.

These things are definitely going to to muddled up with the real things, hopefully not in the official archive, but definitely on the interwebs. People will be using content aware fill to "fix" these pictures, and they'll be sticking them up on pinterest.

But wait, real research doesn't get done on pinterest, so we're OK, right? Yeah, well, real researchers also don't go painting over bits of history with bullshit like content aware fill to see what will happen either, do they? Oh, wait, yes they do. Of course real research is done on pinterest. People with loads of funny letters after their names pull random shit off the web all the time. And even if they don't, they do look at wikipedia, which is the distillation of random shit pulled off the internet.

It's probably not a huge deal. So someone makes a wrong deduction about how dresses were made in the 1930's based on some garbage photoshop invented in some picture. Big deal. Well, this is how history dies. So, yeah, in a way, big deal. It's one cut of a thousand, but I'm still not in favor.

No comments:

Post a Comment