We have so far preserved more than 1.9 million “ephemeral experiences” – exposure to short-lived content that impacts people and then disappears, leaving no trace – that Google and other companies are able to use to shift opinions and voting preferences, and we expect to have captured more than 2.5 million by Election Day.
In emails leaked from Google to The Wall Street Journal in 2018, Googlers (that’s what they call themselves) discussed how they might be able to use “ephemeral experiences” to change people’s views about Trump’s travel ban. The company later denied that this plan was ever implemented, but leaked content (including multiple blacklists) and startling revelations by Tristan Harris, Zach Vorhies, and other whistleblowers show that Google is indeed out to remake the world in its own image. As the company’s CFO, Ruth Porat, said in a November 11th, 2016 video that leaked in 2018, “we will use the great strength and resources and reach we have” to advance Google’s values.
Since early 2016, my team has been developing and improving Neilsen-type monitoring systems that allow us to do to Google-and-the-Gang what they do to us and our children 24/7: to track their activity, and, specifically, to preserve that very dangerous and persuasive ephemeral content.
Since 2013, I have been conducting rigorous controlled experiments to quantify how persuasive that kind of content can be. I’ve so far identified about a dozen new forms of online manipulation that make use of ephemeral experiences, and nearly all these techniques are controlled exclusively by Google and, to a lesser extent, other tech companies.
These new forms of influence are stunning in their impact. Search results that favor one candidate (in other words, that lead people who click on high-ranking results to web pages that glorify that candidate) can shift the voting preferences of undecided voters by up to 80 percent in some demographic groups after a single search. Carefully crafted search suggestions that flash at you while you are typing a search term can turn a 50/50 split among undecided voters into a 90/10 split with no one knowing they have been manipulated. A single question-and-answer interaction on a digital personal assistant can shift the voting preferences of undecided voters by more than 40 percent.