Maybe it’s “Democracy dies in the edits”: Does the The Washington Post have memory holes?

Tonight, I planned to write something complimentary about The Washington Post. WaPo published a brief but excellent piece of investigative reporting by Peter Whoriskey (@PeterWhoriskey) on the organic dairy industry, specifically scrutinizing a company called Aurora Organic Dairy: Why your ‘organic’ milk may not be organic.

Update: As of Monday, May 8th, the image in question has been restored. AFAICT, no explanation has ever been given for either its disappearance or its reappearance.

The gist of that article is that Aurora, whose organic milk is sold in private-label packaging by behemoths like Costco and Walmart, appears to be violating USDA regulations around organic certification. As per USDA rules, if a dairy producer wants to claim that milk from their cows is organic, then they are supposed to let their cows graze (i.e. walk around a grassy field and eat living plant biomass) when the grass in their area is growing. The WaPo piece states that reporters visited Aurora’s biggest dairy complex, referred to in the article as its “High Plains dairy complex”, multiple times in 2016 and saw very little grazing going on, but lots of cows chowing down in feedlots, and also that a satellite photo from last summer (said photo was not included in the article, alas), looked at the pastureland at High Plains and saw mere hundreds — out of the more than fifteen thousand kept there by Aurora — munching grass.

Most damning, however, were three graphs which displayed the results of laboratory tests. Peter Whoriskey, or one of the other contributors to the article or perhaps his editor, sent samples of several brands of milk, including Aurora’s (in the form of some Walmart’s Great Value organic milk), to Virginia Tech and had assays done, looking at the levels of three fatty acids that can serve as signatures for grass-fed versus non-grass-fed milk: linoleic acid, α-linoleic acid (ALA), and conjugated linoleic acid (CLA). In organic milk, levels of the first are lower than in regular (i.e. industrial-style production) milk and levels of the latter two should be higher. The three charts showed the amounts of each, in grams per hundred grams of fatty acids isolated from the samples, and Aurora’s milk comes out looking a lot more like regular milk than organic milk. The Aurora sample was much lower in CLA or ALA than the other organic brands tested and higher in plain linoleic acid. Whoriskey’s piece is great. It’s informative. A reporter and/or contributors have done genuine legwork and obtained new-to-me and (this part is crucial) actionable information about something that could conceivably affect my own and my loved ones’ daily lives. I might decide to buy a different brand of milk and switch brands when it comes to other dairy products the next time that I visit the supermarket. Granted, it’s not earth-shattering, my-world-just-turned-upside-down stuff, but it’s real and relevant.

It also came close on the heels of my critical take on a transcript of a presentation given by The Washington Post’s director of product about the organization’s use of software agents (i.e. “bots”): What it looks like when the wheels fall off; Or, Where’s the good-journalism bot?. The DoP led with high-minded talk about bots not misinforming the public and then immediately switched gears to describing a bunch of real-world, in-use bots that ranged from irrelevant to news consumers — e.g. a deadline nag bot for WaPo reporters — to blatantly harmful to the public, like the one which (on a daily basis) tasked readers with dumbing down their views of a subject until they could be described by one of five emoji s and subsequently showed them which emojis everybody else had chosen. One of the bots, built for internal use only, tracked the amount of attention a newly-published article was receiving and tried to predict how well it would do, engagement-wise, so that the article could be altered if it wasn’t earning whatever was viewed as a sufficient amount of shares, retweets, and the like. That raised alarm bells for me.

Anyway, remember how I said that Whoriskey’s story is great? It is, but it was better yesterday than it is today. It has been edited, without so much as an editor’s note added after the fact. This evening, I was preparing to add a note to the top of my WaPo bots lament praising Whoriskey’s reporting and allowing that, maybe, Bezos’s blog wasn’t completely moribund. Loading the article in my browser and re-reading it, I noticed that the three bar graphs which had made so significant impression on me the day before had vanished without a trace. There was no editor’s note at the end of the story explaining why the charts had been removed. Only one of the most recent reader comments on that page mentions the changes made to the piece: I’m very disappointed that the graphs and the photo of the VA tech dairy science prof were removed from this article. You and me both, buddy.

I’ve reached out to Whoriskey on Twitter, through a reply to his tweet about the article’s publication and a re-tweet of that tweet, and will update this post if I hear back from him.

Off the top of my hand, I can imagine plenty of prosaic explanations for the removal of the charts. For instance, it’s possible that there was an error in the spreadsheet-to-graphics pipeline and they’ve been withdrawn for correction. Or that a flaw was discovered in the lab tests’ methodology or in the handling of the samples. The screencapped comment above, however, also claims that a Virginia Tech scientist’s photo has been removed since they first read the story. I can’t recall seeing such a photo, but it may have been there. Or it may have been excised by an even earlier edit. In any case, while it’s disquieting in and of itself that The Washington Post is heavily modifying stories post-publication, what should deeply concern us all is the fact that they’re doing altering their journalistic output silently, without making even a pretense of notifying readers. Thus, my reference in the title of this post to memory holes.

Orwellian gloom and doom aside, since I really would like to be able to take this information into account when making future dairy product purchases, I hunted down a copy of the image containing the graphs. For the moment, at least, it’s still present on WaPo’s servers. Here’s what it looks like:

Imagine Wikipedia articles without the View history tab linking to a complete, reverse-chronological list of revisions made with a usernames or IP address associated with each. That may be where we’re at, right now, with journalism and yesterday’s newspapers of record. Perhaps it’s an inevitable outcome when you have news organizations which have grown accustomed to churning out multiple versions of the same content for different platforms and for different audiences’ tastes while growing progressively more desperate for clicks and clinging by their fingernails to fading perceptions of legitimacy. If changes to articles are rare and only undertaken to correct or clarify, then there’s little or no shame in notifying your readers of changes to your reporting. If, on the other hand, frequent tune-ups become commonplace and the tweaking isn’t being undertaken for altruistic reasons but rather chasing engagement (e.g. it’s a case of ALLO, PUNY HUMAN! BOT SEZ STORY NEEDS MOAR VIRAL!), then noting changes can get awkward for organizations claiming to be trustworthy sources of real news.