Audio tweets with porn sounds are posted on Twitter. The social network is not ready for their moderation

12 months ago

The social network Twitter began to test the function of audio messages and ran into a number of problems for which it was not ready

As you know, Twitter is actively struggling with the dissemination of untruthful information and sensitive content. In November 2019, Twitter tightened its moderation policy.

Posts with “sensitive materials” fall under moderation. This photo and video depicting scenes of physical violence, accidents, human discharge, the so-called “adult content” (this includes images of sexual intercourse (including pictures and animation), as well as photographs of naked people).

Those sharing this content are encouraged to make their profiles private. To view posts that social network algorithms have labeled as “delicate materials,” users need to further confirm their decision to familiarize themselves with them.

June 17, Twitter began testing a new feature. IOS device users were able to publish on their pages not only text, photos and videos, but also audio messages lasting up to 140 seconds. However, the social network was not ready to moderate sensitive content in the audio recording format.

Already on the day of the test launch of this function, users began to use it to record, for example, porn sounds. One of the users in the comments under the post of the official social network account showed an example of such an audio message and asked: “Is this what y’all want?”

https://twitter.com/Danny_Walterr/status/1273327649134383120?ref_src=twsrc%5Etfw%7Ctwcamp%5Etweetembed%7Ctwterm%5E1273327649134383120&ref_url=https%3A%2F%2Fbykvu.com%2Fru%2Fbukvy%2Fv-twitter-postjat-audiotvity-so-zvukami-porno-socset-ne-gotova-k-ih-moderacii%2F

There may be more problems with auditing: expert opinions
Journalists of the OneZero project interviewed experts in the field of content moderation and ethics. Experts shared their views on the situation and suggested solutions to the problem.

Evelyn Dowk, a teacher at Harvard Law School who studies speech regulation on the Web, noted that this feature in itself is neither bad nor good. However, the negative consequences of its use will be detected in any case (and we are talking here not only about audio support).

“Here, like on any new platform for content on the Internet, all bad things will be present – these will be hateful statements, misinformation, threats, bullying. We know that now it is part of the Internet, ”she said.

Casey University of Colorado’s assistant professor Casey Fisler, whose studies focus on digital ethics and law, also highlighted that adult audio content that is not labeled “sensitive” is not the only problem that the new feature will bring.

Professor Fisler, Professor Aaron Jiang, also pointed out that porn sounds are just one of many abuses on the platform.

Posts where people, for example, simply scream loudly into the microphone can become no less common. According to him, “traditional” abuses (such as posts with insults or harassment) can also go into audio format. The expert stressed that hearing hate words is much more unpleasant than reading them, therefore, the authors of such messages should be called more severely.

The specialist added that another potential problem could be the appearance of audio recordings, the listening of which can be traumatic.

“What if someone uploads audio about someone committing suicide?” It would be very traumatic for users and moderators, ”he said.

The expert also pointed out that it is easier to deal with sensitive content in photo and video format. For example, if you blur a “delicate” image. In this case, the user will have an approximate idea of ​​what awaits him, and based on this will decide whether he wants to look at it.

A similar solution is possible for audio content. One possible solution, according to Jiang, could be to automatically transcribe audio messages and the ability to preview the text. But this decision, the expert noted, is not universal.

“But this is far from an ideal solution. On the one hand, voice recognition technology makes mistakes, it is biased and not so good at recognizing speech patterns and accents of women, non-white people, people who do not speak English. Some sounds may be difficult to translate. If you are really focused on fighting harassment or hate speech, you might want to do an automatic transcription. But how will you transcribe porn sounds or screams? ”Jiang said.

Another problem associated with audio answers is their inaccessibility to deaf and hard of hearing people. Blake Nitko, a deaf, working coordinator of social networks at the National Technical Institute for the Deaf at the Rochester Institute of Technology, noted that he did not consider the idea of ​​transcribing audio messages to be ideal. He condemned the social network for the fact that this problem, like others, will be solved after the launch of a new function.

“You can say that there will be automatically generated signatures. This technology is far from perfect and is not a viable solution. I fully understand the desire to be innovative and find ways to change the format on social networks … But you absolutely must make sure that your platform is fully accessible, BEFORE you release a new feature, ”says Nitko.

Against the background of the active discussion of the above on the Web, representatives of Twitter issued a statement in which they admitted that they could not solve all these problems.

Twitter indicated that they are considering introducing transcription for audio tweets. The company is also working on the introduction of additional monitoring systems for sensitive content for its subsequent labeling. They, as promised by the company, will appear before the function becomes available to all users.

Leave a Comment