W(h)ither the Christchurch Call?

Last week, in conjunction with the UN General Assembly, the leaders of the Christchurch Call community met in New York for the third Leaders’ Summit to discuss progress to date and future initiatives. The Christchurch Call was initiated in May 2019 by New Zealand Prime Minister Jacinda Ardern and French President Emmanuel Macron, following the livestreaming of a gunman killing 51 worshippers at two mosques in Christchurch, New Zealand. It is a partnership of 55 countries, the European Commission, 12 online service providers (including Amazon, Google, Meta, Twitter, and Zoom), the Council of Europe, and UNESCO that aims to “eliminate terrorist and violent extremist content online.” Ardern has also bound New Zealand to “exercise a degree of global leadership on wider digital and tech issues beyond the specific scope of the Call.”

It is apposite to consider after three years what the Call community has achieved so far and what can be expected in the future. The communiqué released by the co-chairs claims internet users are now “far less likely to encounter terrorist and violent extremist content online,” but recent evidence from Buffalo, New York and Texas hasn’t exactly shown a reduction in actual terrorist incidents or the streaming of them. It remains unclear how the link between viewing web content and real-life behaviors operates.

It is clear, however, that banning the online content of only some providers necessarily leads to that content shifting to either non-Call providers or offline entirely, where it is unable to be observed, predicted, or prevented using web tools. The Call leaders’ response is to extend their reach, but this will only ever be an imperfect, ex post response, as both online innovation and the ingenuity of terrorists run ahead of them.

This raises questions about a major initiative announced at the summit: a partnership between New Zealand, the US, Twitter, and Microsoft to engage with algorithm research company OpenMined to “understand the impact of algorithms on users’ online experiences.” The initiative is meant to develop and test “ground-breaking privacy-enhancing software infrastructure” that “could be used to overcome barriers to researching how algorithms drive individuals towards terrorist and violent content” to achieve a “free, open and secure internet.”

As algorithms are responsible for curating and amplifying the content one receives on internet platforms, at first glance it might seem the initiative will shed light on how the relevant algorithms work. However, as it involves only two internet platform companies, it cannot plumb the depths of the algorithmic minefield that already exists. Indeed, the Call’s September 2022 release contains the telling words: “If successful, these technologies will be made available to the Christchurch Call community.” It remains unclear whether this is possible at all.

A deeper examination reveals the true proposal: a secure software platform on which data from multiple companies can be stored to enable researchers to access it without breaching the multiple privacy and data access agreements each company has with its customers. The platform is allegedly needed to reduce the costs and complexity of independent access to the data to multiple firms simultaneously so they can study major algorithms and their effects (and perhaps other things). The initiative expands on an earlier undertaking between OpenMined and Twitter.

The algorithm initiative is therefore a feasibility study and pilot to combine data across multiple platforms to facilitate research. On one hand, this is likely an improvement on the status quo, in which researchers are typically confined to researching the effects of algorithms within one platform. On the other hand, the initiative will not necessarily be confined to algorithmic research. Also, its relevance will be confined to only those platforms that opt to participate; thus, any research cannot be exhaustive or comprehensive.

A further concern is the potential that, in the quest for more comprehensive research data, use of the platform will be made mandatory, or the vast majority of internet firms will opt to join voluntarily. The research will be more comprehensive, but the software operator will acquire considerable market power. This raises significant questions about the governance of the entity managing the enterprise.

So far, governance details are sparse. New Zealand and Microsoft are providing finance, and New Zealand will coordinate with the Call community and work with the US to ensure project outcomes are developed and tested for different types of online service providers. Twitter will provide proof of concept for the infrastructure, and Microsoft will enable its testing on a different platform. Yet there is no transparency about what the software aims to achieve and how it will be tested.

What is clearer is that the scope of the project has implications that stretch beyond simply keeping terrorist content off the internet. Any algorithm used for any purpose on any participating platform will potentially be examinable. Is this really where the Christchurch Call should be heading?