In recent weeks, just take a tour on Instagram to see some stories with a link and a fairly clear request, despite the English: “Ask me an anonymous question”. The responsibility for these contents lies entirely with Ngl, an application that, in reality, is based on a not-so-recent idea: we can only be truly honest when the person we are addressing does not really know who we are.
The operation of Ngl – acronym for Not gonna lieI will not tell a lie in English – in fact, it is quite simple and not so different from predecessors like Ask.fm. With one difference: it is designed to work very well within Instagram. When you download the application, available for Android and iOS, you sign up by entering your username from the Meta social network. At this point, the system generates a link, which can be shared as a link within the Stories or inserted in the profile bio.
Social
Instagram uses facial recognition to verify the age of users
by Simone Cosimi
From the application, the recipient can access a box where they can view what they have received, without obtaining information on the sender. There is also a paid version which, according to Ngl, provides some more clues to find the users who send messages, whose name is not revealed anyway.
In the last month, the application has literally taken off. At the time of writing, it is at number 1 of downloads in Italy on the iOS App Store in the category Styles and trends. Worldwide, it was downloaded more than 7 million times in June alone, according to Apptopia.
The risk of cyberbullying
Along with success, problems could begin for NGL. Almost all applications dedicated to anonymous questions and messages have in fact faced the same kind of difficulty once they reached a large audience. In summary: if it is true that we are sincere when no one knows who we are, is it not equally true that this anonymity could make harassment and bullying easier?
It is in fact precisely this accusation, a few years ago, that led to the center of the news Ask.fm, the platform that in many ways comes closest to NGL. The application, which met with great success in the early 1910s, was even accused of having helped drive some teenagers to suicide, particularly in the United States and England. The issue concerned – and still concerns – content moderation: messages arrived without any filter, leaving some users overwhelmed by waves of hatred and insults.
Does Ngl Content Moderation Work?
It is precisely on this playing field that Ngl tries to differentiate itself from its predecessors. “To ensure the safety of our users – reads the official website -, we use one of the best artificial intelligences for content moderation. Our algorithm can detect the semantic meaning of emojis and ours web scraper extracts specific examples of contextual use of emojis. This means we stay up to date on trends, understand the jargon and know how to filter out malicious messages. ” In other words, Ngl uses a particularly sophisticated AI system, capable of blocking, on the basis of an interpretation not only of the words but also of the context, potentially dangerous messages, which in this way are not even read by the recipient.
The point is: does this strategy work? According to an NBC article, not always. According to the test conducted by the US newspaper, the system is able to identify particularly vulgar expressions without difficulty, but it is not equally effective for formulas that have to do with the bodyshaming or bullying properly understood, such as “you are a loser”, among those that can be mentioned. And this using the English language, the one on which presumably the system has a higher level of training.
In Italian, according to a short test we conducted, the situation is even more problematic. In our language, the system was unable to recognize and block neither obscenities nor bodyshaming nor formulas that have to do with bullying. At the user’s disposal, in cases like these, there is only one button, at the top right of the message display screen, which allows you to report the content and block the sender.