Deepfakes are everywhere these days, or at least anxiety about them. These “synthetic media” videos, which MIT Technology Review cautions will “threaten to further blur the line between truth and fiction,” leverage advances in artificial intelligence and raw computing power to produce videos of popular figures saying things they have never said and doing things they have never done. Previously available only to well-funded digital graphics outfits but increasingly available to the Internet’s broader demos, the ability to create such videos has produced embarrassing problems which our era of unbridled social media has (unfortunately) made entirely predictable.
Watching well-made deepfakes like the one below by Jordan Peele has made me reconsider some ancient epistemological problems and elements of the rhetorical tradition concering character and persuasion. As this video demonstrates, deepfakes present a compelling new case study in the philosophical problem of “indiscernibility,” where the “vividness”–the enargeia or evidentia–of a sensory experience cannot alone serve as sufficient justification for believing what we see or hear:
Ancient treatments of the “indiscernibility” or “indistinguishability” problem of course never envision the example of deepfake videos, but they instead rely on more pedestrian examples like eggs and twins. Here’s Sextus Empiricus:
For in the case of things that are like in shape, but that differ in terms of what is underlying, it is impossible to distinguish the apprehensive appearance from the false and the non-apprehensive one. For example, if there are two eggs exactly alike, and I give them to the Stoic one after the other, will the wise person, after fastening upon them, have the capacity to say infallibly whether the egg he is being shown is a single one, or the one and then the other?
Sextus Empiricus, Against the Logicians I.409 (trans. Bett)
Watching these videos with colleagues and friends, I’m struck by how often we rely on notions of character rather than technical expertise to tell the true from the false. Rather than watch for telltale signs of artificiality like the absence of blinking (akin to botched shadows in photoshopped images), people watch the above video and proclaim that “Obama wouldn’t say that sort of thing.”
These reactions have made me reconsider the nature and aims of “character-based appeal” (or ethos) in classical systems of rhetoric. One of the standard three methods of proof–the others being logical (rooted in logos) and emotional (rooted in pathos)–this appeal to character often gets introduced to students as a method of making an audience receptive to a speaker’s views. (A good example of this method is the captatio benevolentiae, a common tactic in Ciceronian oratory.)
But by positioning character not simply as a matter of winning favor but as a heuristic tool for media’s legitimacy, deepfakes prompt us to radically reevaluate the purpose and function of character appeals in ancient rhetoric. If we remain unable to surmount the Sextus’ epistemological skepticism when it comes to on-line videos (not just eggs and twins), we may have to rely on imperfect metrics like expected speech patterns and moral reputation to tell the true from the false, the doctored from the #nofilter.