I borrowed a phrase in week eleven and moved on. Gabon, 2019: an authentic presidential address dismissed as a deepfake, cited a week later by officers attempting a coup. I called that the liar’s dividend paid out with no fake in circulation, credited Chesney and Citron in the sources, and kept walking. A term borrowed in passing deserves a proper reading, and this one has had a busy seven years.
The phrase comes from Robert Chesney and Danielle Keats Citron’s 2019 article in the California Law Review, a long and careful survey of what synthetic media would do to privacy, democracy, and national security. The liar’s dividend is a short passage inside a much larger argument, and it has comprehensively outrun the rest of the paper. The basic claim is the one usually quoted: once convincing fakes are known to exist, a liar can dismiss authentic evidence, and the dismissal is more plausible than it would have been a decade ago.
That is the version in circulation. The more interesting claim is the second one.
The sharper claim, and the one that almost never survives the citation, is about when the dividend grows. Chesney and Citron argue that it grows as the public becomes better informed about deepfakes. The more successfully you teach a population that synthetic media is convincing and cheap, the more credible you make every subsequent denial by everyone caught on tape.
Part of the dividend comes from the technology. Part of it comes from warning people about the technology.
Sit with the shape of that for a moment, because it is unusual. An awareness campaign assumes that a harm shrinks when you educate the public. This one has a term that grows. The standard defensive reflex, running public education about the existence of fakes, is also a deposit into the liar’s account. That is why a seven-year-old law review passage still repays a close reading.
For most of those seven years it remained a well-argued conjecture. What has changed is that somebody tested it. Kaylyn Jackson Schiff, Daniel Schiff, and Natália Bueno ran five survey experiments with more than fifteen thousand American adults, presenting hypothetical politician responses to stories describing real scandals. They separated two strategies a politician might use. The first invokes informational uncertainty: that recording is fake. The second encourages oppositional rallying: my enemies fabricated this to damage me. Both raised support for the politician. Both did so across partisan subgroups rather than only among the politician’s own side.
I want to be careful about what that establishes, because week thirteen committed me to applying the same discipline to my own sources that I apply to everyone else’s. These are survey experiments using hypothetical responses. They demonstrate that the strategy can pay under controlled conditions with a large sample. They do not measure how often politicians actually reach for it, or what it earns in a real campaign against a real press corps. That is a meaningful limit, and I would rather name it than let a strong citation do more work than it can carry.
Held to that boundary, it is still the most important development I know of since 2019. The dividend has moved from a plausible mechanism to a measured effect, and measured effects are what let an argument be wrong in public.
Which leaves the question of what to do, and here the literature contains an unresolved tension. Josh Goldstein and Andrew Lohn, writing for the Brennan Center in January 2024 as part of a series run with Georgetown’s Center for Security and Emerging Technology, conclude that there is no single answer and propose three levers: establishing norms against making these false claims, developing provenance technology, and enhancing public discernment.
Set that third lever next to Chesney and Citron’s second claim and the two do not sit comfortably together. If awareness feeds the dividend, then improving public discernment is either the solution or a contribution to the problem, depending entirely on what discernment is made of. Teaching people that fakes exist is not the same as handing them a way to check. The first raises doubt and inflates the dividend. Only the second gives doubt somewhere to land.
That distinction is why provenance is load-bearing and detection is not, which is the argument I made in week eleven and am now more confident about: mark the real rather than hunt the fake.
As for where the dividend sits in the rubric, it is a denominator attack and a remarkably efficient one. It adds nothing to volume, frequency, charge, coordination, or mimicry. It degrades resilience directly, by eroding the standing capacity of an audience and its institutions to establish that a thing is real. And it is among the cheapest denominator attacks in the equation, because it requires no production whatsoever. You do not have to make anything. You only have to say the word fake and let seven years of public education do the rest.
Next week: what Finland actually did, which is not what the story says it did.
— J.W.B.
Note 14 of 24