IMAGE: Ayelet Keshet — 123RF

Using machine intelligence to fix human nature

Perspective is the algorithm announced by Jigsaw, the Alphabet ideas incubator that applies technological solutions to fight censorship, cyberattacks and extremism, and is able to detect insults, disqualifications, personal attacks or toxic language on threads.

This is a machine learning algorithm: comments collected from publications such as The Economist, The Guardian, The New York Times and Wikipedia (the experience with The New York Times appears in this article) were evaluated by a number of people through surveys and classified on a scale ranging from “very toxic” to “very healthy,” with toxicity defined as “a rude, disrespectful, or unreasonable comment that is likely to make you leave a discussion.” What the algorithm does is try to recognize those comments, evaluate them, and look for similar cases to assign them a similar label based on probability percentage.

From there, the tool — for the moment available only in English — can be used by anyone who wants to use it to continue training the algorithm. Site managers can use it as they see fit: from simply labeling comments based on their supposed toxicity to help moderation to connecting it with systems that eliminate or prevent certain comments from being sent according to their content, while continuously providing feedback learning.

The algorithm tries to solve one of the oldest problems of internet participation: the fact that people and moderation systems are not scalable. On my own page, which usually publishes just one article a day, I have spent a lot of time and effort moderating comments to try to maintain a healthy and productive conversation, so imagine the task facing a newspaper that publishes a hundred times more articles all kinds of controversial topics such as sport or politics: it is quite simply beyond the possibilities of most media, and has led many publishers to simply not allowing comments, or just abandoning threads to their fate and allowing them to become a troll grotto. This is an important issue, because it threatens to do away with one of the most interesting advantages of the internet: the possibility of participating in forums that can enrich our opinion or broaden our points of view.

It is too soon to know if Perspective will turn out to be a game changer, or if it will fall prey to malicious users who try to poison it by adding false information. At the moment, English words can be introduced on the website and can be added as an API to pages that want to use it for testing purposes. In any case, whether it works or not, the idea of ​​using an algorithm to solve a problem of human nature that people have not been able to solve through minimally scalable solutions is nothing if not a paradox and food for thought…


(En español, aquí)