Twitter Tests New Sensitive Content Appeals Process as it Continues to Refine its Detection Efforts

This will be welcome news for Twitter users who feel that they’ve been unfairly punished by Twitter’s visual identification algorithms for perceived content indiscretions.

Today, Twitter is launching a new test of an updated appeal process for sensitive media violations, to make it easier to query such rulings.

The process will provide a more direct way to action content appeals from tweets, helping users to get a response from Twitter if their content has been mistakenly tagged. 

Twitter’s image recognition processes have been the cause of various high-profile incidents of misclassification on this front. Just last month, the platform came under fire after an image of a badly burned Army veteran saluting in uniform on July 4th was flagged for including ‘potentially sensitive content’.

Realistically, Twitter needs to lean into AI identification as much as possible, in order to limit the distribution of harmful content, and its systems, Twitter says, are leading to improvements in this respect.

In its most recent Transparency Report, Twitter reported that impressions on violative Tweets – those that break Twitter’s rules for whatever reason – accounted for less than 0.1% of all total tweet impressions within the period, with its AI identification tools helping to significantly limit exposure of such over time. Twitter also reported 322% increase in the number of violations of its sensitive media policy, and a 194% increase in the number of accounts actioned for non-consensual nudity.

Overall, these are good results, but that increased activity, and reliance on automated, preemptive restriction, will also lead to more false positives as well.

Which is why this improved appeals process makes sense, and will hopefully enable Twitter to continue optimizing its detection systems, while also catering to users who’ve been mistakenly penalized as a result.

The initial test of the new appeals process will be made available to a small group of users in Mexico, the US and Japan, the latter two being Twitter’s biggest user markets. 

“If you’re in the experiment and haven’t marked your own media as sensitive in settings, you can appeal a Tweet that’s been labeled from the Tweet itself (web only) or label notification.”

It’s a good update, which aligns with Twitter’s broader shift towards increasing protections for users.

#Twitter #Tests #Sensitive #Content #Appeals #Process #Continues #Refine #Detection #Efforts

Leave a Reply