In the past, Wirtualna Polska had built their own internal live content moderation team and used third party systems to moderate the incoming stream of comments. However, the changing legal and political environment in Europe, as well as growing extremism and polarisation in society, meant that these tools are often inadequate for modern journalism. Many publishers have chosen to close their comments sections altogether, as moderation is increasingly an economic burden they can’t handle.

The tragic death of the Mayor of Gdańsk Paweł Adamowicz in 2019 was the impulse for us to start the project of developing an automated moderation system at WP,” explains Dorota Cyndecka from WP.pl.

“This situation showed us at WP what scale of emotions can pour out from users and how quickly we should respond to the flood of especially negative comments.

We were – and continue to be – also motivated by the ongoing observation of the public debate in Poland and the willingness to play a part in improving it. Hence the launch of the #stophejt (stop hate) campaign during the 2019 parliamentary elections in Poland. The awareness that readers wanted to express their emotions and opinions, especially those who do not have space for it on a daily basis, is a positive external factor that has a huge impact on the sense of mission and responsibility of WP.”

Speaking on the partnership, Cyndecka comments, “Our cooperation with the Google News Initiative is very fruitful for two reasons. First of all, it set us up for the design process stages and mobilized us to present and share the effects at regular intervals. Secondly, the Google News Initiative team was helpful with substantive advice thanks to AI experts.

Because Wirtualna Polska are committed to providing a safe space for open discussion they decided against closing their reader comments. Instead, they used Google Cloud tools to create an automated, AI-based system and ease the burden on content moderators. Wirtualna Polska created a Comments Management Platform (CMP) using Google Cloud tools which allowed them to analyze, calculate and convert higher volumes of data at much higher speeds. Wirtualna Polska had not previously used AI technology or machine learning models in this way or on such a scale.

“Readers are more likely to speak up in a safe environment that we achieve through better content moderation,” adds Cyndecka. “Thanks to models adapted to the type of content, people have a better experience amongst others and on the site itself.

Cyndecka concludes, “As the largest information portal in the country, we believe that we can set trends. We share our knowledge from the project at industry conferences and engage socially by creating campaigns such as #stophejt. We believe that these activities will help reach others with good practices.”

Wirtualna Polska plans to develop their project further with the Google News Initiative. The next step will be a system that checks the accuracy of the information provided in the comments — something like a watchdog system. The publisher’s aim is to build an open and secure discussion system that is also reliably factual.