How MeFi (and other providers) deal with trust & safety issues

November 21, 2021 12:09 PM

MetaFilter and several other online service providers responded to a survey about how they prevent and respond to harmful content. A new research paper analyzes those trust and safety techniques: "These findings have implications for policy debates over the regulation of online service providers’ anti-abuse obligations and their use of end-to-end encryption." It's research and it's written in academic language but I figured that it would make sense to mention it here since it's a published paper that analyzes MeFi. See the appendix for some quotes from MetaFilter management.

From the "Limitations" subsection of the "Methodology" section: "while some small entities responded to the survey (such as Lobste.rs and MetaFilter), the data is biased toward bigger organizations with large user bases."

posted by brainwane to MetaFilter-Related at 12:09 PM (4 comments total)

Yes! This was interesting. I think I was responsible for basically 100% of the actual response text that shows up in the paper, though I chatted about it with the team some before and during the questionnaire process.

The note below-the-fold here is definitely pertinent: the framing of the study toward larger social media spaces made answering some of the questions a little bit complicated in terms of interpretation, and I ended up clarifying at least one point after the fact with one of the researchers because it was hard to fit some of the assumptions in the question framing to a site like MeFi.

It's both gratifying and sadly unsurprising to see the degree to which we provided thorough, detailed answers to the questions contrasted against the larger organizations; we don't have a PR department constraining and editing answers to straightforward questions the way larger, more conservative businesses might.
posted by cortex (staff) at 1:55 PM on November 21, 2021 [12 favorites]

I was really happy to see that Metafilter provided thoughtful answers. It was interesting to see that the other site providing detailed responses was Lobste.rs, which is also very small (1 admin, ~2k active users).

The key point I found interesting was that in all but a few categories, user reports really seemed like the highest-value signal for finding abuse.

Also, the paper’s author Riana Pfefferkorn does really fascinating work and is worth following on Twitter if that’s a site you use.
posted by learning from frequent failure at 2:51 PM on November 21, 2021 [3 favorites]

Rank 37,760

Movin’ on up.
posted by rip at 4:38 AM on November 28, 2021

Haven't heard of this lobste.rs site before....
[/me clicks]
oooh, a professional white background!
posted by kaibutsu at 3:42 PM on November 29, 2021 [3 favorites]