Introduction
"Liar's dividend" has become a common phrase in coverage of deepfakes, and it is often used loosely to mean any harm that fakes do. The two law professors who coined it meant something narrower.
This reading goes through the passage where they introduce the term, one sentence at a time, to pin down what the term covers and what it leaves out.
Locating the Passage
The article is "Deep Fakes: A Looming Challenge for Privacy, Democracy, and National Security," by Bobby Chesney and Danielle Citron, published in the California Law Review in 2019. Both were law professors at the time, Chesney at the University of Texas and Citron at the University of Maryland. The journal's website carries the full text free, and the link is in the References.
The article has no numbered paragraphs, so this reading cites it by heading. The passage is a short subsection titled "The Liar's Dividend: Beware the Cry of Deep-Fake News." It closes the article's survey of harms to society, in the second of the article's main parts. The article's introduction also previews the idea in one sentence. All quotations are from the article (Chesney and Citron 2019).
"Deep fake" is the authors' spelling. They use it for fabricated video or audio that is realistic enough to pass as a genuine recording.
Walking Through the Passage
Step 1: Read the sentence that names the term and identify who benefits
The naming sentence comes at the start of the subsection's fourth paragraph: "Hence what we call the liar's dividend: this dividend flows, perversely, in proportion to success in educating the public about the dangers of deep fakes."
A dividend is a payout to someone who holds a stake. The person paid here is in the name: the liar. The subsection's opening has already said which kind of liar. Every harm the article listed before this point came from using a fake "to convince people that fictional things really occurred." Then the authors turn: "some of the most dangerous lies take the form of denials."
So the beneficiary is a person who denies something. This person hasn't made a fake and may never have seen one.
Step 2: Read "things that are in fact true" and note that the harm is to true evidence
The article's introduction previews the idea by saying that deep fakes make it easier for liars to avoid accountability for "things that are in fact true."
The phrase tells you what is being denied. The liar's target is a true report or a genuine recording. In the subsection the authors describe the liar "denouncing authentic video and audio as deep fakes."
This reverses the usual picture of deepfake harm. In the usual picture a false thing is taken for true. Here a true thing is taken for false, or at least placed in doubt. The damage falls on real evidence and on whoever relied on it: a journalist, a court, a member of the public trying to work out what happened.
Step 3: Read the elaboration that the dividend grows with public education about fakes
The third paragraph of the subsection explains why the denial works: "liars aiming to dodge responsibility for their real words and actions will become more credible as the public becomes more educated about the threats posed by deep fakes."
The authors mark this with the word "Ironically." In the naming sentence they use "perversely." Both words flag the same point. Public education about fakes is meant as a protection, and it is also what makes the denial believable.
They then state the result plainly: "a skeptical public will be primed to doubt the authenticity of real audio and video evidence." And they add: "This skepticism can be invoked just as well against authentic as against adulterated content."
The phrase "in proportion to" in the naming sentence sets up a relationship of degree. A public that has never heard of deepfakes gives the liar nothing. A public that has heard a great deal gives the liar a ready audience.
Step 4: Ask what follows for media literacy teaching
The passage doesn't tell teachers what to do. The implication can be drawn from it.
If the lesson people take away is "any recording might be fake," the dividend is at its largest, because that belief is exactly what the liar needs. The authors' sentence about skepticism being "invoked just as well against authentic" content describes doubt that doesn't discriminate.
A different lesson is "here is how to find out whether this recording is genuine": who published it first, whether the original file exists, whether other recordings or witnesses agree. Someone taught this still doubts. They also have a way to resolve the doubt, so a bare denial has less force with them.
The passage supports the first half of this directly. The second half is an inference from it, and the authors don't make it in this subsection.
Step 5: Test the idea against one hypothetical case of a denied recording
Here is an invented case. A city council member is recorded at a private dinner promising a contract to a donor. A local reporter obtains the audio from someone who was present and publishes it. The council member says, "That isn't me. Anyone can make a voice clip with AI now."
Check it against the passage.
- Is the recording genuine? Yes. That matches "things that are in fact true."
- Is the lie a denial? Yes. Nothing fictional is being asserted to have occurred.
- Does the denial lean on public awareness of fakes? Yes. "Anyone can make a voice clip with AI now" is an appeal to what the audience already believes about the technology.
- Did anyone make a fake? No.
All four answers fit. The case would be a use of the liar's dividend whether or not the denial succeeds. How well it works depends on what else supports the recording: the witness, the reporter's account of how the audio was obtained, and whether the original file can be examined.
Key Considerations
The article dates from 2019. It predates the tools that now generate images, voices, and video from a typed description, and it was written about video and audio. The authors were reasoning about where the technology would lead. The term has since been applied to images and text, which goes beyond their wording while following their logic.
The subsection also places the dividend within a wider trend. The authors write that it "would run with the grain of larger trends involving truth skepticism," and they mention the use of "fake news" as a way to wave off damaging reports. They don't present it as an effect of technology alone.
The common mistake is to use "liar's dividend" for any harm a deepfake does. A fraud carried out with a faked video call is a deepfake harm, and a fabricated clip of a candidate is a deepfake harm. Neither is the liar's dividend, because in both a fake is made and believed. The term applies only when real evidence is denied.
A second point to keep straight is that the passage is an argument. It offers reasoning and no measurements. Quoting it establishes what the term means. It doesn't establish how often such denials succeed.
Summary
The passage defines a harm that needs no fake: a liar's gain from the public's knowledge that fakes are possible. A plain-language definition, with one case that fits and one that doesn't:
The liar's dividend is the advantage a person gains by calling genuine evidence fake, at a time when the public knows convincing fakes can be made. It grows as public awareness of fakes grows, and it doesn't require that any fake be made or believed.
Fits: An official recorded making a real promise to a donor says the audio was generated by AI, and some listeners accept the denial because they know such audio can be made.
Doesn't fit: An employee sends money after a video call with fraudsters who used a fabricated image and voice of his manager. A fake was made and believed, so this is fraud by impersonation.
References
- Chesney, Bobby, and Danielle Citron. 2019. "Deep Fakes: A Looming Challenge for Privacy, Democracy, and National Security." California Law Review 107: 1753.