How one anonymous chat app uses AI to filter out trolls and terrible people

How one anonymous chat app uses AI to filter out trolls and terrible people

The website PostSecret, started in 2005 on Blogger, helped kickstart the idea of anonymous sharing. The website was hugely popular, but when the PostSecret app released in 2011 it quickly crashed and burned. The main reason? The same anonymity that brought the appeal gave people free reign to be toxic, terrible human beings. The app quickly flooded with terrible “secrets” and the app was shut down only three months after its release.

If the likes of Reddit and 4chan are any indication, when given anonymity some people will use that to be horrible to each other. Sure, most people know how to handle the responsibility, but as the saying goes, a few bad apples spoil the bunch. One new anonymous chat app wants to alleviate many of the issues that arise when people are given anonymity.

candid convo tnw

Candid, which recently launched on both iOS and Android allows users to post almost anything into different sections, including politics, religion, television, and many others. You name it, and there’s probably a section devoted to it. Having used the app since its beta launch, and after its official release, this writer was impressed by just how nice people were being to each other. As it turns out, that isn’t entirely accurate, and instead many of the hateful things are simply filtered out by the use of AI.

By using a natural language processing system, Candid is able to autonomously monitor everything posted to the app. If it is deemed racist, hateful, or inflammatory it is purged from the application.  Moreover, the AI system will move off-topic posts to their correct section in order to reduce clutter from people trying to interject off-topic subjects into high traffic sections.

candid examples

Candid keeps each user completely anonymous by switching their screen name in every topic they participate in. Your name in one post might be “WildMongoose” and “CuriousSwan” in the next. You’ll never know who you may be talking to from one post the next, but the app rewards users with badges that help identify the type of person you are interacting with. Badges like “Giver” and “Gossip,” and more negative ones like “Hater” are all possible, and are rewarded based on the same AI system.

While censoring content in an anonymous app might seem contradictory, it is sometimes necessary. By removing toxic, hateful content Candid is helping to ensure they will not come under fire like another anonymous app, Yik Yak, who recently came under fire due to a harsh, uncalled for bullying incident.

This post is part of our contributor series. It is written and published independently of TNW.

This post is part of our contributor series. The views expressed are the author's own and not necessarily shared by TNW.

Read next: The ATL-based startup, Shofur, launches new bus line in Austin