Vous êtes sur la page 1sur 4

John Schlichting - 7/27/2018

Ethics of Video Impersonation


Ethan Brauer - Philosophy 1332

Understanding the relationship technology holds in our fast-paced developing society is


crucial to benefit from its implementation. Without proper oversight, infrastructure, and adoption
of society many new technologies are subject to failure. Emerging technologies have numerous
unintended consequences for consideration. This paper will examine the morality of
impersonation and deception regarding two emerging technologies, Voco and Real-time Face
Capture. Using both, users can create realistic videos of any public figure saying anything given
correct source input. The moral questions driving this paper are: When would impersonation or
deception be permissible? Given these technological capabilities, do the benefits outweigh the
costs? How might abuses of these technologies deteriorate trust in video evidence or news? We
will examine these questions in the specific application of social media and political campaigns.
The first technology is called Voco, it is a text-to speech (TTS) software. Adobe debuted
Voco as an audio editing and generating software in 2016. With twenty minutes of source audio
of the desired figure, the software can create the remaining phonemes of English not present in
the sample. The software uses a convolutional neural network (CNN) to interpret the inputted
waveforms associating each wave with each phoneme. It outputs the user’s desired sentence by
restructuring these waveforms to create each desired word. This lets the user generate new
sentences of audio that the sample never contained. Developer Zeyu Jin demonstrated this
unsettling ability by changing a joke the conference host Jordan Peele had made, he reversed
word order and added new words. The uncanny resemblance of the generated audio and Jordan
Peele’s voice demonstrated Adobe’s efforts to perfect TTS capabilities. Mr. Jin stated that they
have researched watermarking, detectability, and forgery prevention as ways of addressing these
concerns. Pragmatic applications apply to movies, music, or radio industries where actors make
errors that need to be re-recorded, which costs money. This technology lets these small
expensive mistakes be changed with ease creating a better quality product for the consumer. The
second technology, Real-time Face Capture, generates video of a target figure with desired
motions by using two video feeds, actor and target feed. Software records and tracks points on
the actor’s face. The target feed is then matched with these points and the generated facial
expressions of the target are changed to match the actors. These tracked points create a modified
face template of the target. One approach that has shown successful video generation is called
dense face tracker developed at Stanford University.1 Their approach uses every inputted pixel to
generate a fine template, while other systems, like Snapchat filters, use sparse features to create a
rough template. This technology would allow directors to refine every detail of their movie,
developers to create realistic in-game motion, and researchers a way to record facial movements

Zollhöfer, Michael. “Face2Face: Real-Time Face Capture and Reenactment of RGB Videos.” ​CVPR 2016
1

Publication​, Stanford University, 2016, web.stanford.edu/~zollhoef/papers/CVPR2016_Face2Face/page.html.


accurately. This software presents the opportunity fabricate videos; while this has always been a
possibility the software had been limited to video editors. This Real-time Face Capture operates
with a webcam, an actor, and a target video. The ease and widespread availability of such a
product would allow users to generate any type of desired video evidence.
In a political setting, it is a politician’s duty to represent their constituents’ beliefs and act
accordingly within the law to reflect and safeguard these beliefs. The political setting provides
the most readily avenue for abuse and unjust outcomes for two reasons. First, politicians are
inherently recorded both audibly and visually on a consistent basis for extended periods of time
with most of these recordings made publicly available. This gives easy access to the inputs
required by the two technologies, being audio and visual source material. Second, in recent
elections fake news has been shown to alter voter opinion2 by exploiting our social media sharing
and inability to fact check with the pace of content generation. This affords politicians
deniability for statements as any previously condemning video evidence could be argued to be
fake. This also allows rumors or fake stories to hold some semblance of validity without
verifiability; which misinforms, deceives, deteriorates trust3, and ultimately harms society. In this
context, it is morally impermissible to make use of these technologies as doing so treats rational
humans, the voting citizens, as mere means, allows for abuse of duties inherent to the political
setting, and has harmful outcomes on society’s trust.​ ​Without a specific disclosed contract like
those present in the film industry between actor and producer, any person that uses these
technologies without proper disclosure is inherently treating other rational beings as mere means.
The target of the fake news is unable to consent to her image’s alteration and the consumer of the
fake news is unable to know of her misinformation/deception. These instances seem harmful and
abundant with much more drastic outcomes than a bad video edit. Our world is a closely-knit
hive of social media consumers, we share and post with an egotistical hunger related to our
“fears of missing out” and self-promotion. The dynamic of social media and fake videos allows
for abuses of society and overall distrust in media. Specifically, circular reporting in a political
setting. The practice where source A publishes misinformation that source B republishes, A then
cites B as their original source. This has led to misinformation becoming repeatedly obscured
and erroneously cited. In our democracy, the media has often been called the 4th branch;
operating as a check on power and providing accurate information on our world and government.
Studies from PEW show that 62% of adults receive news from social media and that “viral” fake
news is more widely shared than mainstream news4. These statistics demonstrate “attention

2
​Guess, Andrew, et al. “Selective Exposure to Misinformation: Evidence from the Consumption of Fake News
during the 2016 U.S. Presidential Campaign.” ​Dartmouth University,​ European Research Council, 9 Jan. 2018,
www.dartmouth.edu/~nyhan/fake-news-2016.pdf.
3
​Edelman Intelligence. “2018 Edelman Trust Barometer Executive Summary.” ​Trust Barometer​, Edelman
Intelligence Consultancy, 20 Nov. 2017, www.edelman.com/trust-barometer/.
4
Gottfried, Jeffrey, and Elisa Shearer. “News Use Across Social Media Platforms 2016.” ​Pew Research Center's
Journalism Project,​ PEW Research Center, 26 May 2016,
www.journalism.org/2016/05/26/news-use-across-social-media-platforms-2016/.
economy”, an idea that attention is a resource of which a person only has a limited amount. The
goal is to engage with the user for as long as possible. Fake news stories effectively engage the
user by either running counter to preconceived notions or acting as an echo-chamber. A study of
European engagement with fake news5 found that despite false news sites having relatively little
reach outside of their engaged group; these small engaged groups exceeded the level of posts and
shares (measures of engagement) seen with popular news sites. This idea to draw attention and
engage the user directly relates to political influence as well. Political campaigns rely on social
media to spread their message, their information, and gain voter support. As a result,
(demonstrated in our recent election, though apparently debated) fake news has a large impact on
public perception. A falsified video of a politician in a relevant attack ad jeopardizes a
politician’s chances of election.
Objection​ -- This technology is permissible because a practical method of verification
and disclosure solves the ethical problems faced. A politician will still be held accountable for
the statements she makes as her video could be verified as being made of her own volition. The
ability to distinguish between a falsified video and a real video still allows for treatment of
rational agents as means, but not as mere means because pertinent information is present and
verified. It would also promote trust in social media as sources of content would be validated.
While there should be safeguards for political campaigns to reduce misinformation, if we operate
within the law, it is our 1​st​ amendment right to say what we want regardless of misinformation.
The specific counter example in this political setting is the element of satire. This plays a crucial
role in what makes the 4​th​ branch of democracy work so well. Political satire offers a flexible
option to address political topics in a scrutinizing light while retaining humorous appeal to the
consumer. Social media posts and other publications using falsified videos are already utilize
political satire; Alec Baldwin impersonating President Donald Trump on SNL is a recent
example. Satirists highlight societal injustices by using humor to exacerbate the obviousness of
the problem.
The practical method described is not practical; the verification the objection refers to is a
limited process and can be an insurmountable goal if the original user falsifies/generates the
video in a certain manner. The verification relies on analyzing metadata embedded in the video
file, however this data can become obscured and difficult to interpret once the video has been
reuploaded and repackaged multiple times.6 Which occurs each time a video is shared, uploaded,
posted, or spread. Which makes this method almost unable to be applied practically within the
confines of social media and the current state of technological verification. We know, that it is

5
​Fletcher, Richard, et al. “Measuring the Reach of ‘Fake News’ and Online Disinformation in Europe.” ​Factsheet -
Reuters Institute​, Reuters Institute - University of Oxford, Feb. 2018,
reutersinstitute.politics.ox.ac.uk/sites/default/files/2018-02/Measuring%20the%20reach%20of%20fake%20news%2
0and%20online%20distribution%20in%20Europe%20CORRECT%20FLAG.pd
6
​Supasorn Suwajanakorn, Steven M. Seitz, and Ira Kemelmacher-Shlizerman. 2017. Synthesizing Obama: Learning
Lip Sync from Audio. ACM Trans. Graph. 36, 4, Article 95 (July 2017), 13 pages. DOI:
http://dx.doi.org/10.1145/3072959.3073640
Alec Baldwin and not Donald Trump talking in the video. The abuse comes down to intent.
Political satire intends to impersonate a target while informing the audience of an issue. Video
falsification intends to recreate a target, whereby responsibility for statements made are
accountable to the target by concealing information regarding the source of the video. Political
satire inherently assumes the audience is aware of the joke or knows of the impersonation. This
is the reason Mr. Baldwin exaggerates aspects of President Trump’s character, to appeal to the
audience and create the sensation that he is performing a comical routine. Whereas, a fabricated
viral video conceals the intent to impersonate and creates the sensation of validity. It is important
to note that freedom of speech should not be oppressed, but that limiting the use of these
technologies does not infringe upon the 1st amendment. If anything, preventing widespread use
of this technology causes less infringement because statements cannot be as readily attributed to
fake news and accountability in our political system is preserved; fostering trust in information
sources and a better relationship.

Vous aimerez peut-être aussi