Andrew Betts describes a recent browsing session:

I tapped a link in the Twitter app, which showed as…, got a page in Twitter’s in-app webview, where the visible URL bar displays the reassuring But this is actually content from Russia Today, an organisation 100% funded by the Russian government and classified as propaganda by Columbia Journalism Review and by the former US Secretary of State. Google are allowing RT to get away with zero branding, and are happily distributing the content to a mass audience.

This is not OK. This is catastrophic.

Betts is talking about content cached by Google’s AMP (Accelerated Mobile Pages) platform. While the goal of the platform is ostensibly to speed delivery of pages, it also serves those pages from a Google URL. With the URL spoofed, the origin of the content is hard to discern. This “recklessly devalues the URL,” Betts writes, and makes AMP an attractive petri dish for fake news:

If the world’s biggest content discovery and delivery platforms prioritise security, performance and popularity, over authenticity, evidence and independence, well, the likely result is an exponential rise of simplistic, populistic thinking, inevitably spreading and amplifying until false beliefs become tacitly accepted as facts. … [W]e need a much stronger focus on authenticity as a strong ranking signal. This is not only critical to avoid potentially huge societal implications of bad decision making, but also cultivates better content by improving incentives for creators.

Totally agree. As our answer machines continue to be overwhelmed by propaganda (and worse), they need to listen for new ranking signals. We need to build systems smart enough to know when they’re not smart enough—and that know to complain when their immune systems have been compromised.

Technology decisions in AMP are affecting far more than page speeds, aggravating what I consider to be one of the big civic crises of our times: the erosion of trust in the fourth estate. At the very least, let’s protect the URL as citation and origin model.

Read more about...