Who Thrives in a World of Deepfakes and Misinformation?

Source: cottonbro / Pexels

Artificial intelligence-altered videos and images are exponentially rising, with some estimating that deepfakes online are doubling every six months.

AI advances have made making videos and audio of events that have never occurred easier. People no longer need expensive and sophisticated technology to create synthetic media. Synthetic media is any content (e.g., text, images, videos) that has been fully or partially created using AI.

The Liar's Dividend: How Deepfakes Are Changing the Way We View Evidence

This climate of powerful generative AI has brought about a phenomenon called the "liar's dividend," which describes the benefit to those who claim that anything is fake, even objective evidence.

In a world of AI-generated videos and audio, the liar's dividend benefits people who use this technology to dispute and raise skepticism about objective evidence; in other words, a strategy to deny reality.

The term "liar's dividend" was coined by law professors Bobby Chesney and Danielle Citron in 2018 and describes the strategy of casting doubt on objective evidence as fake or manipulated. This phenomenon is one of misinformation about misinformation and benefits those who weaponize public skepticism and take advantage of a climate of uncertainty as a means to escape accountability.

In recent studies, researchers have found that politicians who use this strategy to dispute credible evidence raised in a scandal are more able to garner support by disputing text-based evidence. However, this does not work as effectively for video evidence.

This may change as the public becomes more familiar with the potential manipulation of videos and AI-altered videos become more commonplace. Researchers also found that claiming objective evidence was fake is more effective for gaining support than for the politician staying silent or apologizing.

As the authenticity of videos becomes increasingly difficult to prove, the liar's dividend will likely continue to pay off.

In courts, lawyers are already using "the deepfake defense" to cast doubt on video and audio evidence in legal cases, including a recent case involving statements from a 2016 recorded video interview of Elon Musk. The lawyers for Tesla attempted to use the "deepfake defense" to cast doubt on statements made in the video.

The judge later described this strategy as "deeply troubling" and did not condone what was seen as "hid[ing] behind the potential for their recorded statements being a deep fake to avoid taking ownership of what they did say and do." This type of defense will likely be used more frequently, and, as a result, courts and juries will likely expect more evidence of authentication or forensic expert verification of certain video, audio, or digital evidence.

Although the technology is new, the battle between those seeking misinformation and those trying to detect it through authentication is longstanding. In this case, the very AI technology that produces the deepfakes holds the solution to detecting it, creating a tension between deepfake creation and deepfake detection that has been labeled an AI arms race.

Deepfakes and the Erosion of Trust: How to Combat the Liar's Dividend

One strategy to combat the liar's dividend is to continue investing in, incentivizing, and developing more reliable, affordable, and accessible deepfake detection technology. Previous research has found that humans and AI models are imperfect at identifying deepfake videos.

However, humans and AI models make different types of mistakes. This suggests that human-AI collaboration could be a helpful solution to authenticating videos. Stronger and more reliable deepfake detection technology will provide the public with more confidence in verified and authenticated evidence.

Artificial Intelligence Essential Reads

AI's Turing Test Moment

Using AI to Enhance Your Email Communication as a Leader

However, it will not stop people from trying to create skepticism about objective evidence or the authentication technology itself.

A second strategy is increasing public education and awareness of the liar's dividend. This will have the dual benefit of helping people learn how to determine the trustworthiness of claims about videos and also protecting consumers from scams that utilize deepfakes.

As the line between reality and deception becomes increasingly blurred, the expectation for proof of what is real will evolve. The hope is that people can adjust their awareness and learn how to adapt to this new uncertainty.

Marlynn Wei, MD, PLLC © 2023

See my related article
Are Humans or AI Better at Detecting Deepfake Videos