News for hackers. Stuff that matters.

A comment on Nobel scientist's deletion from Wikipedia points to wider bias. Parent comment

Re:Nobel scientist's deletion from Wikipedia points to wider... by cool_dude85 on Thursday August 19, 2021 @01:19PM (#28236602)
>You have to live with the fact that sometimes you're wrong.

You don't have to live with the fact that your institutional process for sending articles to this binary classifier is biased, per the published study discussed in the article. Women's profiles make up 19% of the profiles on wikipedia but 25% of deletion nominations.

Even if your classifier was totally gender-blind (unlikely), by virtue of the process that leads to the classifier being invoked, more women's articles will be deleted.

[ Parent | on HN ]
Re:Nobel scientist's deletion from Wikipedia points to wider... by jhgb on Thursday August 19, 2021 @01:29PM (#28236726)
The problem is that those numbers apply to articles that have already been written. Such small discrepancy could easily be explained by other things, such as editors writing new articles with their own bias. If you then apply even objective criteria to those articles and end up with deleting a larger portion of some group, I don't see how you can claim that the deleting classifier is biased without knowing what was deleted.
[ Parent | on HN ]
Re:Nobel scientist's deletion from Wikipedia points to wider... by cool_dude85 on Thursday August 19, 2021 @01:40PM (#28236867)
"Such small discrepancy"? Women's articles are 30% more likely to be nominated than by random chance. That's actually quite a large discrepancy to be explained away by handwaving about how editors might be biased in including too many non-notable women. The journal article even claims it's exactly the opposite:

"Studies show that women’s biographies are slightly more notable than men’s (Wagner et al., 2016)"

[ Parent | on HN ]
Re:Nobel scientist's deletion from Wikipedia points to wider... by jhgb on Thursday August 19, 2021 @01:56PM (#28237078)
To me, 30% does absolutely not seem like a large discrepancy if there's push for inclusion of more women in Wikipedia and if these efforts are trying to compensate a 4:1 discrepancy in current articles. At that point you'd need several times more women's articles than you currently have so the idea that slightly more poor quality articles may be being deleted is definitely one of the obvious options.
[ Parent | on HN ]