Lydia Laurenson Polarisation and Peacebuilding Strategy on Digital Media Platforms: Current Strategies 11 that Facebook’s primary business interest is keeping users engaged with Facebook, whereas Google’s primary business interest is locating the best resource the user is searching for on the open internet. This difference means that Google is incentivised to support the open internet. Thus, the company has developed many tools intended to keep the open Internet healthy, including free technical improvements for websites operated by outside organisations, and other tools that are open to the public. In 2013, for example, Jigsaw launched Project Shield, a free service to help human rights organisations, journalism organisations, and others protect their websites from malevolent attacks. In 2018, Google launched a search engine for fact checks, which got written up in the journalism trade publication Poynter by Daniel Funke. Funke quoted a Google research scientist stating that “it’s possible that Google will add more signals to its algorithm to surface more fact checks in search” once the fact-checking engine is out of beta — so Google might start actively fact-checking outside websites when users turn them up in Google search. But for now, the new project simply helps users to fact-check any topic they are interested in, if they have the idea to fact-check it on their own. (Like Facebook’s factchecking efforts, Google’s appear to be reliant on outside civil society partners.) If Google starts surfacing fact-checks in Google search, it may solidify a new direction for the company. Like Facebook and all the other major platforms, Google spent many years seeking to avoid the direct arbitration of truth on its platform. But it is unclear that such arbitration can be dodged. For instance, Google’s “featured snippets” — text that appears directly in Google search — are pulled directly from outside websites, which means the snippets are often wrong. Yet those snippets bear the authority and trust of Google’s brand because they appear directly in Google search. Journalists like Sarah Perez at TechCrunch wasted no time criticising this feature, and Google’s first response was to introduce “multi-faceted” featured snippets that show multiple answers from many websites at once (which, of course, could all be incorrect). If Google does add fact-checking to search, that will be a more direct intervention, and a more direct acknowledgment of responsibility. Another example from 2017-2018 is a project called PerspectiveAPI, built by researchers at Jigsaw, an R&D organisation that used to be part of Google and is now owned directly by Alphabet. PerspectiveAPI is an attempt to use artificial intelligence to make Internet comments less toxic. PerspectiveAPI started with the development of automated “toxicity scores” written up in this 2017 Medium post by Jigsaw researchers. Then, in 2018, Jigsaw partnered with Oxford’s Rhodes Artificial Intelligence Lab to experiment with integrating those scores gently into comment sections. In another Medium post, the scholars talk about adding “speedbumps” to toxic comments — perhaps similar in spirit to the “viral slowdowns” Facebook has been experimenting with. These speedbumps operate by intervening before a user posts a toxic comment and asking the user to tone it down, perhaps with a question like: “Your comment is likely to be perceived as condescending by other users, and will be assessed by a moderator. Do you wish to rephrase it?” Much of this is publicly documented, and PerspectiveAPI has also been implemented