As the 2024 presidential election approaches, the evolving role of generative artificial intelligence (AI) in shaping political discourse has garnered heightened scrutiny. These AI systems, such as ChatGPT, are engineered to produce written, visual, or audio content from extensive datasets, enabling the creation of deepfakes—highly realistic but falsified images and videos. These capabilities are raising concerns about their potential impact on the electoral process.
Experts like Thomas Scanlon from Carnegie Mellon University's Software Engineering Institute have highlighted the risks associated with deepfakes, particularly their ability to convey believable misinformation that can be difficult to distinguish from authentic media. These concerns are underscored by the potential use of deepfakes by both domestic and foreign adversaries to disseminate false narratives about political figures or manipulate their speeches, aiming to mislead voters.
Thus far, more conspicuously outlandish AI-generated content, such as a digitally fabricated image of Donald Trump riding a lion, has surfaced in the public sphere. Yet, worries persist that subtler, more deceptive AI-generated content could emerge closer to Election Day. Instances like fake videos inaccurately stating the closure of voting locations could have a tangible effect on voter turnout, due to the time required to address and rectify such misinformation.
The impact of AI-generated content on public trust is of particular concern. An erosion of confidence in democratic institutions and elected officials is a potential consequence, according to the Block Center for Technology and Society at Carnegie Mellon's Heinz College. The increasingly inundated information landscape requires individuals to judge the credibility and value of the information they encounter, a challenge noted by Randall Trzeciak, who directs a Master of Science programme at Heinz College.
The ease with which misinformation can be generated and spread through AI is augmented by its capacity to produce content that resonates with existing beliefs or leverages personal trust within online communities. This ability to tailor misinformation amplifies its potential to deceive, elaborated Ari Lightman, a professor at Heinz College.
Deepfakes possess indicators, as outlined by Scanlon, that can signal their inauthenticity: inconsistent lighting, mismatched reactions, and significant fluctuation in visual quality. Recognising these signs is vital for discerning fact from fiction.
Despite the urgent need for regulation, the legislative response remains fragmented. Federal laws specifically targeting deepfakes are nonexistent, and while some states have moved to prohibit malicious deepfakes, these regulations are not universally applied to election-related matters. A notable legislative effort is underway in Pennsylvania, where a bill proposes penalties for those disseminating AI-generated false impersonations in campaign ads, pending proof of malicious intent. This bill is yet to be debated.
The effective regulation of AI misuses like deepfakes is acknowledged as challenging, with authorities potentially resorting to exemplary enforcement actions. Beyond legislation, Lightman underscores the necessity to address the broader issues of political polarisation and diminishing trust in societal institutions, which contribute to the propagation of misinformation.
These technological advancements present both opportunities and challenges for the electoral process. How society navigates this complex landscape could shape the integrity of democratic processes in the years to come.
Source: Noah Wire Services