By Kyra Theodore
@redlineproject
AI disclosure: ChatGPT was used to assist in brainstorming ideas and organizing the structure of the content. MidJourney was used to create images, and Runway ML was used to create the video. The summary was generated by SummarizeWise custom GPT and the podcast was generated by NotebookLM.
Summary: This paper explores the impact of deepfakes on public trust in media and the broader democratic process. By analyzing cases like the Nancy Pelosi video and AI-generated images of Taylor Swift, it highlights how deepfakes spread misinformation and create distrust, especially when public figures exploit this distrust through “liar’s dividend,” casting doubt on authentic media. This erosion of credibility makes it harder for audiences to discern fact from fiction, undermining democratic engagement. The author argues that, while debunking efforts are necessary, they’re often too slow to counteract viral spread, leaving media in a challenging cycle of detection and credibility preservation.
In 2004, U.S presidential election, a doctored photo of Senator John Kerry sitting beside actress Jane Fonda at an anti-Vietnam War rally began circulating. Despite the photo being an obvious fake and debunked quickly, it spread like wildfire during the campaign. The association between Kerry and Fonda, who was notorious for her anti-war stance, was embedded in many voters’ minds. On election night, a voter told an NPR exit poll interviewer he couldn’t bring himself to vote for Kerry. When asked why, he said he couldn’t get the image out of his head. Even after the interviewer explained the photo was fake, he doubled down, “I know, but I couldn’t get it out of my mind.” In this case, the truth wasn’t enough.
If a poorly photoshopped image could sway voters 20 years ago, what happens in an era where low-cost or free generative AI tools can create fake videos and audio that are nearly identical to reality? As deepfakes improve and become widely available, what if proving something is fake doesn’t matter? In this article, I will look at the complicated role deepfakes play in eroding public trust in media and the consequences for democracy.
Deepfakes: What They Are and How They Work
Deepfakes represent advanced forms of digital manipulation that utilize machine-learning algorithms, specifically Generative Adversarial Networks (GANs). GANs work with two neural networks: the generator and the discriminator. The generator creates fake images, videos, or audio; the discriminator checks them and tries to determine if they are real or fake. As the generator gets better at creating more convincing forgeries, the discriminator gets better at detecting them. This is an iterative process that results in media that’s almost indistinguishable from real content.
To achieve this level of detail GANs process multiple angles and details of the subject. For example, with photos, it focuses on facial features and angles. With videos, it takes a more holistic approach and looks at actions, movements, and speech patterns. The discriminator checks how well the generator captures these subtleties. Then, the two repeat the cycle. This makes the output more realistic each time.
In contrast to older video editing software, modern deepfake tools are freely available, unlicensed, unregulated, and can be used by amateurs (instead of experts) with basic computing abilities and equipment. This democratization has led to an explosion of deepfakes on the internet, ranging from harmless fun to malicious disinformation. Indeed, deepfakes are spreading like wildfire on social media platforms like Facebook, Twitter, and YouTube and can reach millions in minutes before being flagged as fake. This is causing concerns about disinformation and manipulation of public opinion and erosion of trust in digital media.
PODCAST: Listen to an AI-generated podcast about the threat posed by deepfakes to media credibility and democratic processes.
A prime example of this occurred in 2019, when a doctored video of Nancy Pelosi — which made her look drunk and slurring her words–went viral online. In a matter of days, the video racked up over 2.5 million views on Facebook and was also shared by several politicians.
More recently, Taylor Swift’s image was altered in AI-generated photos that showed her endorsing Donald Trump. These pictures originated on Trump’s social media platform, Truth Social. They featured Swift with her fans wearing T-shirts with pro-Trump messages. One of the images had Swift dressed as Uncle Sam with the caption, “Taylor wants YOU to VOTE for DONALD TRUMP.” Trump reposted this and added “I accept!,” effectively giving credibility to the disinformation.
But the problem isn’t just the technology itself. With deepfakes on the rise, the public will become more aware of them and more skeptical of what they consume. This has the potential to erode public trust in the media, leaving voters more likely to reject true information. This creates a scenario where political figures can use that skepticism to further their own agendas.
The Liar’s Dividend: Exploiting Doubt
In addition to generating fake content, deepfakes also enable public figures and individuals to evade responsibility by benefiting from what is called the “liar’s dividend.” Coined by law professors Bobby Chesney and Danielle Citron, the term refers to how bad actors can use the existence of deepfakes to cast doubt on legitimate evidence.
The liar’s dividend directly feeds into the growing trend of calling everything you don’t like “fake news”, a tactic that’s become more and more common among politicians and public figures in recent years. The very existence of deepfakes gives bad actors a way to discredit real evidence by spreading doubt. . The possibility of deepfakes lets people claim real media is fake. This casts doubt on the truth, even when no manipulation has occurred. This is exactly what the broader “fake news” weaponization is all about — to discredit any media or reporting that’s inconvenient or damaging.
Donald Trump’s use of “fake news” is a classic example of this. Throughout his political career he has called negative press fake news and conditioned his supporters to automatically distrust any media that doesn’t fit his narrative. It’s worked because it plays on existing media skepticism and in the age of deepfakes, the claim becomes even more believable. With hyper-realistic manipulated media, no proof is needed to dismiss negative coverage as fake. Just the suggestion it might be altered is enough to create doubt.
The liar’s dividend and the “fake news” accusation feed into each other creating an environment where truth is under attack. Even when media institutions produce factual, verifiable content their credibility is questioned because deepfakes exist. This allows bad actors to dodge scrutiny and responsibility by casting doubt on the evidence presented against them. As a result it becomes much harder for journalists and fact-checkers to do their job as watchdogs of democracy because any attempt to hold powerful people accountable can be dismissed as fabricated or fake.
This phenomenon has even made its way in courtrooms, notably in a lawsuit against Elon Musk. Throughout the trial, his lawyers suggested that a 2016 video of Musk talking about Tesla’s autonomous driving features was a deepfake. There was no evidence to back it up but just the suggestion made people question the video’s authenticity. This is a great example of how deepfakes, even if not used, can undermine factual evidence. We also saw this in the Jan. 6 Capitol riot trials where defendants said the videos showing them at the riot could be deepfakes. While these defenses failed, the fact they were made at all shows how deepfakes can change how evidence is contested in both legal and public spaces.This “liar’s dividend” changes the way we talk to each other, and those in power can dismiss inconvenient truths with ease.
Indeed, evidence no longer speaks for itself; individuals and institutions have to go to great lengths to prove real content is real. Even when they do, skepticism lingers. And the consequences of this skepticism can extend far beyond media or courtrooms. In 2018, after Gabon’s President Ali Bongo had been out of sight for months, rumors circulated that he had died or was incapacitated. In an attempt to shut down these rumors, the government released a video of Bongo delivering a New Year’s speech. Instead of reassuring the public, the video sparked widespread suspicion that it was a deepfake. Citizens pointed to the president’s odd movements and inconsistencies in the footage as evidence it was doctored.
Although the video was later proved to be real, the damage was done. Indeed, the mere possibility that the video was a deepfake led to unrest, widespread distrust of the government and even a coup attempt. The Gabon case exemplifies the issue posed by the existence of deepfakes: seeing is no longer believing. As a result, even real content is dismissed as false because of the ingrained distrust of visual media. In this case the government’s attempt to reassure the public backfired as deepfake suspicions fed into existing doubts about the president’s health.
The Erosion of Trust in Media
Deepfakes have eroded public trust in the media at an alarming rate and it’s not just isolated incidents. A recent study by Weikmann et al., found people who were shown deepfakes had a big drop in trust of audio-visual media overall. This wasn’t limited to the deceptive content itself but to all media formats including legitimate content. This is key to understand because it means deepfakes don’t just deceive us in the moment — they leave behind long term doubt that affects how we interact with and perceive media going forward. Indeed, the study revealed that once you’ve been deceived by a deepfake your ability to trust future media (video, audio, immersive content) is severely compromised. This supports the argument that deepfakes are a threat to media credibility as a whole. In this case, the issue is not just about detecting fakes but about undermining trust in the institutions that are supposed to deliver the truth.
This crisis of trust directly affects democratic processes where informed decision making depends on access to accurate information. An Adobe study revealed that a majority of people in the following countries (84% U.S., 85% U.K., 84% France, 80% Germany) are worried that the content they consume could be manipulated to spread misinformation. And 70% of people in these countries think it’s getting harder to verify if the content they see online is trustworthy. This growing uncertainty creates a vicious cycle where citizens may not be able to make decisions with confidence based on the information available to them, especially in high stakes situations like elections.
Likewise, Weikmann et al. shows that after being deceived by deepfakes participants not only lose trust in the media but also lose faith in their own ability to detect fake content in the future. This loss of self-efficacy is especially dangerous in the democratic process. If people can’t confidently distinguish fact from fiction, especially during elections, the risk of manipulation and confusion grows exponentially. The study by Adobe shows that 80% of U.S. respondents, 78% of U.K. respondents and 70% of German respondents think misinformation and harmful deepfakes will impact future elections. In a democratic society where voters rely on accurate information to make informed decisions, the proliferation of deepfakes undermines the very fabric of the electoral process. When people can’t trust the information they’re exposed to they’re less likely to engage in political discourse and more susceptible to manipulation by bad actors who will exploit these doubts for their own gain.
The psychological impact of deepfakes goes beyond just loss of trust in the media; it weakens the very fabric of democratic engagement. The erosion of self-efficacy creates a public that is skeptical of the media and unsure of their own ability to tell the truth from lies. This is a dangerous combination as it can lead to mass disengagement from political processes and general mistrust of institutions. As more people start to doubt not only the content they consume, but also their ability to evaluate it themselves, the space for misinformation and manipulation gets bigger.
This is supported by the Adobe research findings, which also shows that the rise of misinformation is causing people to change how they use social media. A significant number of respondents (39% U.S., 29% U.K., 37% France, 24% Germany) have stopped or reduced their use of specific social media platforms because of the amount of misinformation they see. This is a clear sign that deepfakes and disinformation are not just theoretical; they’re driving people away from key communication platforms and fragmenting the public sphere.
Combatting Deepfakes: The Role of Media
Deepfakes are designed to deceive and in an era of rampant disinformation media outlets are often tasked to debunk these fabrications. Fact-checking and debunking deepfakes can only go so far as the content spreads faster than the correction. Moreover, A 2020 study from Stanford University’s Internet Observatory found that it’s hard to convince the public that deepfake content is fake after it’s gone viral. Even when fact checkers proved a deepfake video of a political figure was fake, a large portion of people still believed it was real or were uncertain. This was seen in cases like the altered video of the U.S. House Speaker Nancy Pelosi where even after it was debunked millions of people had already seen and shared the manipulated content.
Media faces a tough challenge when it comes to deepfakes. On one hand, we need to inform the public about the existence and dangers of these manipulated videos, images and audio. And on the other, by focusing too much on a specific deepfake, news outlets can make it more viral than it would have been otherwise. Deepfakes are sensational, visual and auditory and that makes them engaging which drives clicks, shares and further dissemination. This amplification effect puts media organizations in a tricky situation: by covering a deepfake they risk drawing more attention to it and giving disinformation more reach than it would have organically.
Making things worse, is that news organizations now have to invest in expensive technology and expertise to verify visual content in real time. AI driven deepfake detection tools are being developed but they are still playing catch up to the technology used to create deepfakes. This arms race between creation and detection puts a lot of pressure on media institutions to keep up. The cost of deepfake detection technology can also be too high for smaller news organizations which may not have the resources to verify visual content fast enough.
FAQ: Deepfakes and the Erosion of Trust
1. What are deepfakes and how are they created?
Deepfakes are highly realistic fabricated media, including videos, images, and audio, created using artificial intelligence (AI). They leverage machine learning algorithms called Generative Adversarial Networks (GANs). GANs consist of two neural networks: a generator that produces fake content and a discriminator that evaluates its authenticity. Through an iterative process of creation and evaluation, deepfakes become increasingly convincing, often indistinguishable from genuine media.
2. Why are deepfakes a threat to democracy?
Deepfakes pose a significant threat to democracy by eroding public trust in media and information. They fuel the spread of disinformation and can be used to manipulate public opinion, especially during elections. The existence of deepfakes allows individuals to dismiss legitimate information as “fake news,” creating an environment where truth is constantly under attack.
3. What is the “liar’s dividend,” and how does it relate to deepfakes?
The “liar’s dividend” describes the phenomenon where individuals, particularly public figures, exploit the existence of deepfakes to cast doubt on legitimate evidence or accusations against them. Even if no manipulation has occurred, the mere possibility of a deepfake allows them to dismiss inconvenient truths as fabrications, benefiting from the public’s skepticism.
4. How do deepfakes impact public trust in media?
Deepfakes have significantly eroded public trust in media. Studies have shown that exposure to deepfakes leads to a decline in trust in all media, including legitimate sources. This is because deepfakes create a sense of uncertainty and doubt, making it harder for individuals to discern truth from falsehood.
5. How do deepfakes affect individual behavior and engagement with information?
Beyond trust in institutions, deepfakes impact individuals’ confidence in their own ability to identify fake content. This “loss of self-efficacy” can lead to disengagement from political processes and a reluctance to participate in online discussions, as individuals become more skeptical of all information they encounter.
6. What challenges do media organizations face in combating deepfakes?
Media outlets face a daunting task in debunking deepfakes. Often, the spread of fake content outpaces fact-checking efforts. Additionally, by covering deepfakes, news organizations risk amplifying their reach. Furthermore, investing in deepfake detection technology is costly, potentially creating a resource gap between larger and smaller media outlets.
7. Can deepfakes influence elections?
Yes, deepfakes have the potential to influence elections by spreading misinformation, manipulating voters’ perceptions of candidates, and undermining trust in the electoral process. The ability to quickly create and disseminate highly believable fake content targeting specific demographics raises serious concerns about election integrity.
8. What can be done to mitigate the threat of deepfakes?
Addressing the threat of deepfakes requires a multi-pronged approach, including:
- Technological advancements: Developing robust deepfake detection tools and exploring methods to authenticate genuine content.
- Media literacy: Educating the public on identifying deepfakes and fostering critical media consumption habits.
- Legislation and regulation: Exploring legal frameworks to hold creators and distributors of malicious deepfakes accountable.
- Platform responsibility: Social media platforms must take proactive steps to prevent the spread of deepfakes and enforce policies against disinformation.
—–
Editor’s note: Students in Mike Reilley’s AI Journalism (COMM 294) Fall 2024 undergraduate course experimented with AI storytelling tools to create these stories, following the AI use guidelines on The Red Line Project’s Principles page.