Navigating our way through the fog of war
Europe offers one path forward, if regulators can get it right
In the wake of the Hamas terrorist attack on October 7th, major digital platforms like YouTube, TikTok, and Facebook have all been scrambling.
I won’t relitigate all the ways in which we’ve already seen platforms dropping the ball in meaningful ways but it is worth noting two key points: (1) Twitter seems to be, by far, the worst and it’s the least surprising thing in the world given all the terrible product & staffing decisions Elon has made in the last year, and (2) as Peter Kafka rightly points out, the fog of war is not a problem that’s unique to social media.
The truth is that the entire world is struggling to stay up to speed on what’s true and what’s not, and if you’ve studied any history at all, you know that it’s a phenomenon as old as time.
But we also know one of the ways forward: transparency and not just transparency, but a very specific kind of transparency.
When the world’s attention suddenly focuses on a single event, it exacerbates all the inherent weaknesses of social platforms and will inevitably bend to the point of breaking even the most robust content moderation operations & policies (or fully expose the inadequate and under-resourced ones). It’s not only that the stakes suddenly matter more than ever but the tidal wave of attention from a public desperate for the latest updates means that good actors and bad actors suddenly flock to the platforms with a frenzy of content. Some of it illegal. Some of it that violates platform policies. But also a lot of it that just falls in between.
When we were at CrowdTangle, some of the busiest & most stressful, but also the most impactful times, we ever had were during moments of crisis.
Whether it was national disasters like hurricanes or typhoons, humanitarian crises where real-world violence had broken out, or in the months and days leading up to an election, we saw the power of helping civil society and the public understand and sort through all the competing, misleading and messy public narratives that were taking hold and doing their best to nudge them in better directions.
Because the truth is that one of the best structural elements of a functional information ecosystem is the ability for it to monitor itself (see Network Propaganda by Yochai Benkler, Robert Faris, and Hal Roberts if you want to learn more, including why the lack of self-policing is one of the biggest weaknesses of the conservative media ecosystem in the U.S.). And we’ve already seen that play out in this conflict. It was the very front-page, public nature of the New York Times’ initial reporting on the hospital bombing (as well as all the other outlets that made the same mistake) that made it possible for the claim to be litigated and ultimately corrected, albeit slowly.
Public stories on social media, however, can be much harder to see. Even some of the biggest ones.
Compared to something like the front page of the New York Times, there is simply way more of it, it’s coming from way more sources and it’s moving much faster. If you want any ability to monitor or respond to it, it has to be intentionally and thoughtfully organized. But to have any hope of doing it at all, it starts with some access to data. Unfortunately, Twitter has basically killed off their public interest APIs, TikTok & YouTube have historically offered very little and there is no CrowdTangle team dedicated to helping journalists and civil societies with crises, including things like the Israel/Palestine fighting.
The good news, though, is that some of the largest platforms in the world are now required, under the Digital Services Act in Europe, to comply with something called Article 40.12, a provision that requires platforms to provide access to real-time public content (and sometimes called the “CrowdTangle provision”).
However, the bad news is that based on early looks at some of the solutions they’re building, they mostly seem to be missing the point of it.
Despite only focusing on public data (that has less but not zero privacy risk), TikTok seems to be focused on putting the data inside tightly controlled data rooms where only a very small handful of qualified academics can access a small subset of public data points with very small rate limits using custom code libraries they’ve written themselves. It is not designed to be user-friendly or broadly accessible. Meta has a slightly better approach, one that theoretically allows for a more diverse group of researchers to get access, (even potentially investigative journalists, according to Meta), but it still seems to be missing the main point. Like TikTok, it’s missing a lot of the most basic, out-of-the-box analytic features that would make it actually useable for civil society in real-time.
For TikTok and Meta, they both seem to be actively improving their solutions, which is perhaps the most promising sign of all but at the moment, the problem is they’re both treating their solutions like research tools instead of monitoring tools. And what we’ve seen in the Israel/Palestine conflict, not to mention the war in Ukraine, and almost every other major civic and political event over the last 10 years, is that when civil society has the ability to monitor the emerging narratives around those events in real-time and study the information ecosystems they exist within, they can play a powerful role in helping guide the public to some semblance of truth during some of our most important moments.
The question going forward is whether regulators will get Article 40.12 right and at the moment, that remains to be seen.
I’m going to write more on all the solutions in the coming few weeks, including Google’s unique approach to the provision and why it might ultimately might be a much better one than Meta and TikTok’s. It’s also worth noting that among the public data points they are making available in their tools, both TikTok and Meta have conveniently decided not to include whether they’ve added a fact-check label to a piece of content but haven’t offered any principles for what data points they’ve chosen to include and which ones they haven’t.
Other transparency & policy news:
Like many others, I was frustrated to see Thierry Breton’s poorly written and poorly timed letters to platforms over disinformation around the Israel/Palestine conflict. For those of us who believe in the DSA, it was a huge setback and only served to confirm the worst fears about elected officials’ heavy-handed use of internet regulation to pursue their own political priorities. DFRLab had a good write-up on it here.
When it comes to TikTok’s first version of their Researcher API, there is a massive gap between the data points they’re offering and the amount of data that is actually publicly available. You can see a comprehensive audit of the difference in this great write-up.
One of my absolute favorite developments in the entire transparency space over the last few months is this scorecard, methodology, and report on how transparent different AI systems are. I highly recommend it.