The Internet doesn’t last forever, for everyone.

News outlets are notoriously bad at saving their work. Archiving the journalism that you pour blood, sweat, and whatever else into is a crucial step to having a lasting impact. But 19 news organizations out of 21 in a study conducted earlier this year weren’t taking any steps to archive their online content (no, saving it in a Google Doc or Github doesn’t count). As I wrote then: “For a field that likes to consider itself the author of the first draft of history, the vast majority of those authors don’t prioritize how to save and share that history in the future.”

But another equally important issue is unpublishing — a news outlet choosing to remove a piece of reporting from the internet, usually on the request of a person mentioned in the story, often related to a crime. There’s the argument for minors not being haunted by an SEO-friendly stupid offense from years ago when they’re applying to college or looking for a job, but there’s also the question of erasing that first draft of history. You don’t know who will be missing that piece of information in whatever context in the future.

There’s quite a gray area between these two points, and Deborah Dwyer has spent the past few years analyzing it. She’s a PhD candidate at UNC Chapel Hill and is writing her dissertation on the ethics and practicality of unpublishing. Drawing from Kathy English’s 2009 study on the same topic, Dwyer revisited the issue with a dozen qualitative interviews and a survey of 100+ newsroom leaders in 2017 and found the same result: Very few news outlets have a stringent policy or general standards. Very few outlets really know what they’re doing with unpublishing.

“We know how people handle corrections — they have a policy. What they don’t necessarily communicate is how they handle unpublishing and removing from the public view,” Dwyer said.

She defines unpublishing as:

The act of deleting factual content that has been previously published online in response to an external request prompted by personal motivations such as embarrassment or privacy concerns.

Here’s some of what she’s found so far:

1. 80 percent of news outlets surveyed had established unpublishing policies, but almost half of those were not in writing and only two percent were shared beyond the newsroom.
2. Even inside the newsroom, the policy wasn’t clear: 14 percent of newsrooms included the unpublishing guidelines in an official staff manual or handbook, a third said people learned about them on the job informally, and 40 percent said they learned of them only when an issue arose.
3. Who has the final say? Usually it is the editor-in-chief and/or the managing editor, but they’re not the ones who ultimately control the website. Yep, the IT team: 64 percent of respondents said IT was capable of removing content autonomously (18 percent were not even sure if the team could or not). About a third said they were unsure if their IT department knew that they did have a policy.
4. When content is removed, it’s not clear where exactly it still is. Less than five percent of respondents said they maintained a system to track the requests and decision-making process for unpublishing content. Forty-eight percent said they don’t have a system or even hold onto the emails asking about unpublishing.
5. Three-quarters of unpublishing requests were related to crimes.
6. Half of newsrooms that do unpublish items update the content with a note about its removal, and some write a follow-up story, especially if it pertains to a court case. (Those charges were dropped or convictions expunged? Here’s a quick post that will show up in search results first.) But about a fourth of them remove the content in question with no reader notification whatsoever. Some also have certain kinds of articles “sunset” or automatically be unpublished after a few years.
7. Anecdotally, Dwyer found that some newsrooms ask unpublishing requesters to provide documentation of the change that is leading to them want the content down, usually legal documents — but others do the legwork themselves or ask a reporter to look into it. Two instances involved people who were at risk of suicide related to the content that they wanted unpublished; one of the newsrooms required certification from a therapist that this was an issue.

Some newsrooms, in an attempt to be more compassionate, have opted out of covering smaller-scale crime or publishing mugshots. If you don’t publish coverage of minor crimes in the first place (yes, after the hypothetical of “if people don’t commit minor crimes in the very first place”), then you don’t have to unpublish it later. But they’ve also formed systems to evaluate their coverage retroactively:

“We started this back in July: If you were charged with a minor crime — we took out violent crime and some other things — and went to the judge and used the system that the taxpayers pay for to get your record sealed, we would take your name out of the stories. You had to provide the proof [that you’d been to the judge] because once it’s sealed, we can’t see the sealing order,” Chris Quinn of Cleveland.com/Advance Ohio told Nieman Lab last year. “It really comes down to: How long does somebody have to pay for a mistake?”

How forgotten is forgotten? Who gets to be forgotten? How do different mediums handle unpublishing differently?

“This is not a technology problem. This is a society problem to answer: how willing are we to put people’s past in context and allow people to change,” Dwyer said. She pointed out that these unpublishing scenarios have all involved individuals thus far; when companies or governments start requesting unpublication, that’s a different story. (This is also a much trickier situation in countries with right-to-be-forgotten laws.)

After studying this for the past few years, Dwyer is now an unpublishing committee convert. But she also cautioned that the people making these decisions need to understand the technology that is involved in unpublishing and also need to be transparent with readers and staff who may wonder someday where a story went.