Twitter offers rewards to those who detect biases in its algorithms

The social network Twitter announced Friday that it will award rewards to users and researchers who detect potential, for example, racial or gender biases, in the algorithms running on its platform.

• Read also: Digital: Comprehensive Attempts to Harmonize GAFAMs

• Read also: Government inaction puts Canadian media at risk

• Read also: Facebook bans users for questionable reasons

It is, according to the company, the first competition in the topic, with prices reaching 3,500 dollars.

The idea is modeled after competitions offered by some websites to find vulnerabilities, Roman Choudhury and Juta Williams, the company’s directors, explained in a message.

“It is hard to find biases in machine learning models, and sometimes companies find that they inadvertently violated ethics only after they are published,” these officials say. “We want that to change.”

In the model developed for vulnerability detection, researchers and hackers alike helped IT security managers “create best practices to identify and manage vulnerabilities to protect the public,” Roman Choudhury and Juta Williams explain.

“We want to develop a similar community” to detect algorithms’ biases.

Twitter in April introduced its ongoing work to make the algorithms behind the scenes of the platform more ethical, a way to respond to criticism of the dangers associated with these technologies.

The social network abandoned its photo-cropping algorithm a few weeks later, after discovering that it was slightly biased in favor of white people.