Category: Rhetoric

ancient deepfakes at classical world

Office Board by John F. Peto

The latest issue of Classical World includes my article on the shifting parameters of successful “impersonation” or sermocinatio in the ancient world, particularly as described in the fourth-century Ars Rhetorica of Gaius Julius Victor. Unlike his predecessors Cicero and Quintilian, who see impersonation as an occasion for “exaggeration and artifice” rooted in the so-called “grand style,” Gaius Julius Victor recommends that speakers use “less overt and even covert means” for taking on the personalities of others. In other words, I trace how impersonation shifts in the fourth century from a method of comedic, rhetorical excess to one of subtlety and even deception.

By recommending a style of impersonation that goes unnoticed, Gaius Julius Victor anticipates our own era’s novel methods of manufacturing deceptive likenesses in deepfake videos, where “concealing one’s fiction behind an expert veneer of sprezzatura is key to the impersonation’s persuasive potency.” His novel stylistic recommendations for sermocinatio in the fourth century show us how an impersonation might be designed to go undetected, not unlike various genres of contemporary disinformation.

The published article is hosted at Classical World, and the pre-print is available through my personal site’s archive.

deep probabilities

In The Atlantic today, David Frum records perhaps the first gunshot in deepfake campaigning:

And then, at 8:25:50 pm ET, the president retweeted an account he had never retweeted before. The account had posted a video of former Vice President Joe Biden, crudely and obviously manipulated to show him twitching his eyebrows and lolling his tongue. The caption read: “Sloppy Joe is trending. I wonder if it’s because of this. You can tell it’s a deep fake because Jill Biden isn’t covering for him.”

As I’ve written here before, deepfake videos are emerging as a new form of political rhetoric. With technological wizardry now available to anyone with a Verizon bill, our visual political commentary is shifting from the ubiquitous political cartoon to the imagery of digital caricature—from the opinion page to the viral tweet. And the ubiquity of the ink pen, an earlier era’s weapon for political cartooning, now competes with the ubiquity of the smartphone processor. Jacob Schulz at Lawfare:

Deepfake creation used to require a serious computer and a good baseline of technological skill. But that barrier to entry has begun to erode. iPhone deepfake apps have arrived, and they’ve made creating deceptive media easier than ever. An iPhone-created deepfake tweeted by an anonymous user with only 60,000 followers received a presidential retweet within an hour of posting. The era of the deepfake apps has arrived.

Schulz writes that deepfake technology is “unlikely to sway the 2020 election.” Instead, he sees a more imminent danger in “cheap fakes—more rudimentarily edited deceptive videos—and clips that simply remove the underlying context of a politician’s comments,” for example a May 2019 video of House Speaker Nancy Pelosi that slowed her speech to give the impression of drunkenness. (Pelosi, like Trump, is a teetotaler.) I hope he’s correct. But even if deepfakes don’t demonstrably sway the current election cycle, their increasing proliferation in the months ahead promises what Frum calls “an experimental test of the rules of social media.” He continues: “Because the account retweeted by Trump explicitly labels its video a ‘deep fake,’ it arguably does not violate Twitter’s anti-deception policy.” Are caricatures like the one in Frum’s essay mere fiction or outright deception? Worthy of a take-down request or protected by the principles of free artistic expression? These questions, of course, bear not merely on literary considerations of forgery and impersonation but also on legal matters like libel.

In his foundational Handbuch der literarischen Rhetorik, Heinrich Lausberg observes that caricature “need not be historically true—it must only be probable” (§821). As the genre of the political video clip slips from documentary evidence to creative fabrication, the language of probability will displace the language of fact and evidence. That’s my prediction, for what it’s worth. Another, to boot: Schulz may be right that deepfakes are not likely to sway 2020, but let’s remind ourselves that 2024 begins mid-November.

on disinformation sprezzatura

In his 1528 Book of the Courtier (Il Cortegiano), Baldasarre Castiglione coins the word sprezzatura, an important term in the history of rhetoric which has no direct synonym in English. Sometimes people translate it as “nonchalance” or “studied carelessness” (OED) or maybe something like “graceful effortlessness.” It reminds me of those German words that require a short paragraph to capture fully. As Castilgione himself describes it, sprezzatura is something more than mere “casualness”—it involves some kind of concealment, constructing a veneer of extemporaneous authenticity for something deliberate and even calculated:

I have found quite a universal rule which in this matter seems to me valid above all others, and in all human affairs whether in word or deed: and that is to avoid affectation in every way possible as though it were some very rough and dangerous reef; and (to pronounce a new word perhaps) to practice in all things a certain sprezzatura, so as to conceal all art and make whatever is done or said appear to be without effort and almost without any thought about it.

Il Cortegiano, I.26 (trans. Singleton)

In Castiglione’s text, sprezzatura is a display of Renaissance virtuosity, an expert command of language and manner that makes mastery look easy. But even if sprezzatura has a five-century history behind it, I’m now seeing traces of it—like so many other rhetorical concepts—all over the much more modern and technological landscape of Internet disinformation.

I was especially reminded of sprezzatura while reading through the Transatlantic Working Group’s report “Actors, Behaviors, Content: A Disinformation ABC,” referenced recently over at Lawfare in a conversation with Camille François of Graphika. Its ‘B’ refers to “Deceptive Behavior,” a topic that refers to Facebook’s ban on “Coordinated Inauthentic Behavior.” What is “studied carelessness” if not a close cousin of “coordinated inauthenticity”? What is the core tactic of disinformation if not convincingly avoiding “affectation” in service of persuasiveness?

The overlap between Renaissance sprezzatura and Internet disinformation becomes even clearer in light of Castiglione’s emphasis on maintaining the illusion of effortlessness:

Therefore we may call that art true art which does not seem to be art; nor must one be more careful of anything than of concealing it, because if it is discovered, this robs a man of all credit and causes him to be held in slight esteem. […] So you see how art, or any intent effort, if it is disclosed, deprives everything of grace.

Il Cortegiano, I.26 (trans. Singleton)

Maintaining one’s “credit” through assiduous cultivation of a plausible identity is central to so many of our digital interactions—see the “fictitious online personas” documented in the Mueller report (vol. I, p. 41). This “careful concealing,” of course, predates the last decade’s disinformation campaigns, perhaps best illustrated in the now-canonical New Yorker cartoon up top, published in 1993. Just as the artful, even seamless obfuscation of identity has always been part of our digital lives, the virtuosic concealment of “any intent effort” has been central to persuasion since at least 1528. Hiding your craft is, as ever, the whole game.