Hacker News Reader

Comment by wpietri | Hacker News Reader

Yeah, I think there's a big difference between unpublishing and retracting. Simply making a story disappear is a common way of hiding a fuckup. (Also, sometimes a way of making a difficult person go away.) Retracting is admitting the mistake.

I'd like this a little better if they had some way to access the old material in a fashion that made it clear it was withdrawn. But I think this is way better than just leaving the old story up in a form that might be harmful to people. I'm strongly influenced here by Wikipedia's "Biographies of Living Persons" standard, which is much stricter than that for articles on other topics: https://en.wikipedia.org/wiki/Wikipedia:Biographies_of_livin...

For those curious, the Wayback link is here: https://web.archive.org/web/*/https://www.npr.org/2018/04/03...

Edit: Actually, user Buge found the real archive link here: https://web.archive.org/web/20180403204501/https://www.npr.o...

View on HN

There are four options here

1. Leave the story unchanged

2. Publicly retract the story, but keep the text available with a disclaimer.

3. Publicly retract the story and "unpublish" it, i.e., remove the text from the website.

4. Silently unpublish the story.

Everyone agrees that #1 and #4 are very bad. ColanR is claiming that #2 is better than #3, and in particular that #3 conflicts with existing NPR policy. We are inferring NPR policy from an informal blog post, not some official document, which was written in the context of third parties requesting that content be taken down. However, it is quoting an AP Editor policy that is not in that context.

You can argue that NPR didn't intend to endorse this policy in the context when they were the ones who messed up, rather than it being a complaint involving a third party. But then it seems rather hypocritical.

I feel you are correct.

But also wanted to add another option between 2 and 3 where I feel a more valuable editing could also be done.

2.1: Publicly retract the story, and edit the previous text (with visible or not removals/edits) along with a disclaimer, to improve the content and present a fairness assessment of it's text for new and previous readers.

I've encountered this before, with both visible (highlighted or striked off content) or completely deleted parts with some "erroneous/unverified content deleted" kind of indication, and I really appreciated the transparency an understanding why something was reviewed.

I feel that would be a valuable approach to these kinds of publications.

Yes, I agree. One way to do this, which works equally well for stories with minor corrections as those that are fully retracted, is to just make the entire edit history publicly accessible. Then the "up-to-date version" of a full-blown retractions would be the retraction notice and explanation, while leaving the original version(s) of the article available.

The problem with this approach is that search engines and other bots have no eye for context. Keeping inaccurate content online, even with a retraction up top, can result in confusing search results and even the propagation of the original inaccuracy. NPR can make efforts to make the correct content appear to search engines and other bots, but it can't guarantee that search engines or bots will get the right net content. When there's a question of fairness, it seems to me it's best to remove the potentially unfair content, especially when it's so potentially inaccurate.

Are you worrying that people will do a Google search, see a result that says "Retracted Story: Man Bites Dog", and then go on repeating this as if it's true? Like, of course, that's possible, but they could get false information from a million other places. Anyone can lie on the internet.

Personally, I'm not worried about a human reading it, but I am worried about a machine reading it and ingesting "facts" about me because they don't know what a retraction is.

Of all the ways that machines can and will misinterpret information on the internet, stories labeled "RETRACTED" are the least concerning.

Yeah, but do the guys making ML models at the credit rating agency or immigration agency know or care about that? I think the parent poster has a point here, who audits these models and their systematic errors (biases)?

Imagine John Doe was named as the owner of the site, and he is not in fact connected to it at all. When people search his name, the top 20 results are this incorrect fact.

No one is repeating it like it were true, but the man's reputation is harmed and he could sue. Maybe he did sue in fact.

There is this simple trick called a redirect that solve the problem.

I think the issue here, and perhaps the deciding factor between 2 & 3 is that the original story names the "alleged author" (basically of the villain of the story) and my read of the retraction is that they can't confirm that the person named as the villain is, in fact, actually the bad guy.

So you can have option 3.5 and keep it published with names removed to protect the "could be innocent" or retract the entire story as they've done.

Your intermediate solution of wiping just the names is not necessarily my preferred solution (I'm honestly undecided), but I endorse it as strictly better than removing the entire text from the internet.

Reply on news.ycombinator.com