While we welcome assessment of hate speech prevalence, we are concerned that
@twittersafety
’s latest analysis is not methodologically robust. Combined with Twitter’s refusal to engage with third-party scrutiny, this appears to amount to a whitewash.
We recently partnered with
@Sprinklr
for an independent assessment of hate speech on Twitter, which we’ve been sharing data on publicly for several months.
Sprinklr’s AI-powered model found that the reach of hate speech on Twitter is even lower than our own model quantified 🧵
@TwitterSafety
1. This study uses slurs as a proxy for prevalence of hate speech, removing those apparently used in non-hateful contexts.
Much hate is conveyed without typing common slurs, including veiled/targeted harassment & abuse, images/videos, deliberate misspellings, rarer slurs, & more.
@TwitterSafety
While no method is perfect, blunt approaches oriented entirely around the use of keywords will miss things, and cannot be the sole solution employed by platforms to moderate harmful activity.
@TwitterSafety
3. Details of the methodology have not been released. There is therefore no capacity for independent scrutiny. If Twitter is confident that these findings are robust and that they indicate improvement, they should have no issue in allowing third party assessment.
@TwitterSafety
4. It is important to measure impressions and engagement to understand how far hate is spreading. However this is not the whole story. For example, hate targeted at a single person requires only 1 impression to have impact. This is another way Twitter’s methodology is incomplete.
@TwitterSafety
5. Independent analyses have found that hate speech on Twitter is becoming more prevalent. This is also the felt sense of many on the platform. Most recently, ISD and CASM have shown a sustained doubling in antisemitism using a robust methodology.
@TwitterSafety
We are calling on Twitter to release the methodological details of this study for independent assessment, and to resume good-faith engagement with third parties to understand and tackle hate on the platform.
@ISDglobal
@TwitterSafety
Agree, if
@twitter
feels the findings are valid then share. It seems many people have left Twitter but those stats would show where if people do not close their accounts (inactive)? Is hate amplified due to decreased members? Racist/hate tweets seem to be at record levels.
@ISDglobal
I have reported hundreds of accounts openly posting slurs and threats. To this day they are still very active. Communities full of some of the most vile comments ever seen. I could spend hours reporting them, and still only scratch the surface of how many are contributing.