ChatGPT and similar AI-driven chatbots have swiftly embedded themselves into users' everyday lives across diverse backgrounds, outpacing even popular platforms like Facebook and Instagram in adoption rates. While AI models like ChatGPT offer substantial benefits, they also carry risks that may threaten national security and peace. Concerns arise around misinformation, whether generated by the model itself or used to craft misleading content. With the 2024 U.S. presidential election approaching, worries about potential misuse of this technology are increasing.
But are these concerns justified? Is this fear of AI’s influence rooted in reality, or is it merely exaggerated by those less informed about the technology's capabilities?
Indirect Influence
When considering ChatGPT's impact on elections, some might envision it directly swaying voters' opinions. However, the influence AI chatbots wield is often indirect, stemming more from their ability to produce convincingly human-like content in various formats and local dialects, making it harder to distinguish from genuine human communication.
This ability to produce localized, believable content represents one of the greatest risks of these AI chatbots. Unlike before, there’s no need for local personnel to create misleading information, as the AI can generate high-quality, abundant content quickly. This makes it highly suited for dissemination across platforms such as Google and social media networks like Facebook and Instagram.
External parties have also recognized ChatGPT’s potential to produce persuasive, human-like misinformation. In one instance, an Iranian group used ChatGPT in an operation called “Storm-2035” to generate misleading content targeted at Latino communities in the United States to sway votes. OpenAI addressed this incident in an official statement, explaining how ChatGPT generated misleading content based on false inputs and refined it to resemble authentic human communication before its release on social media.
According to OpenAI, accounts involved in this operation were banned from accessing ChatGPT, and the content generated through this operation did not reach a level of widespread influence that would directly endanger users. Nevertheless, the possibility remains that undetected accounts may continue to circulate higher-quality, more widely accepted misinformation.
The Era of Customized Misinformation
Misinformation content is typically designed for broad appeal, often lacking the cultural nuances needed to resonate with specific demographics. However, ChatGPT effectively addresses this issue, quickly customizing misleading content to suit diverse target groups. For example, in the "Storm-2035" operation, ChatGPT adapted misinformation to reach Latino or Hispanic communities in a more targeted way.
ChatGPT’s ability to either broaden or narrow the scope of targeted misinformation is a powerful feature. It allows for creating highly specific content, like targeting women over 50 in the southern U.S. who enjoy conspiracy-themed dramas. This targeted approach can have a more substantial impact compared to generalized misinformation.
A New Era of Data Collection and Targeted Advertising
In March 2018, the Cambridge Analytica scandal emerged, revealing how Donald Trump’s campaign leveraged targeted advertising to appeal to specific audiences based on their social media data and preferences. While the controversy focused on data access, the use of personal data itself was not the core issue. AI's ability to swiftly analyze massive amounts of online content and user accounts now makes real-time, tailored data delivery even easier.
AI-powered predictions can also increase influence during elections, as noted in research by experts from Cornell University, Google, MIT, and Harvard. According to the study, AI’s analytical power enables companies or political groups to forecast election results with greater accuracy. This can inform strategies designed to sway outcomes, allowing AI to indirectly influence election results.
Building Misleading Chatbots
While generating targeted misinformation is concerning, an even more potent threat lies in the possibility of deceptive chatbots spreading misinformation. This approach involves either compromising existing chatbots or creating new ones that disseminate false information to users.
Though hacking into platforms like ChatGPT or Gemini is challenging, external entities could develop their own misleading chatbot and promote it long before elections. With the right preparation, this strategy could pose a greater risk than simply publishing misinformation on social media. It offers a more dangerous path toward influencing public opinion in election campaigns.
Why Did Trump Outperform in Charisma in the Last Three Days According to the "Sentiment Analysis" Page?
A Comparison Between Trump and Kamala Harris
As the U.S. presidential election draws near, global media coverage of the candidates has increased. It appears that Donald Trump’s personality and charisma have received more positive media attention recently, compared to his competitor, Kamala Harris, who faced more negative coverage.
Using the data dashboard on the "Sentiment Analysis" page, developed by experts with the help of Al Jazeera, the difference in positive and negative news coverage between the two candidates is clearly evident. The dashboard shows a total of 183 news articles published between October 30 and November 2 discussing the personalities and charisma of the candidates. Of these, Trump received 47 positive articles, accounting for 25.68%, which was the same percentage as the negative articles, which also totaled 47. On the other hand, Harris faced 36 negative articles, which made up 19.67% of the total, compared to only 15 positive articles (8.20%), with the remaining articles being neutral.
When delving deeper into the news coverage through intensive Google searches, Fox News emerged as the leading platform, with a significant number of articles discussing charisma and personality, with Trump dominating this coverage. Harris, by contrast, had fewer articles focused on her personality (41 articles), with 8.46% being positive, 23% negative, and the rest neutral at around 68.5%. Although other outlets also discussed the personalities of both candidates, Trump’s coverage stood out.
Trump gained much positive coverage recently, particularly following statements from the daughter of the late basketball legend Jim Brown, who expressed solidarity with Trump’s campaign. In contrast, Harris benefited from endorsements by basketball star LeBron James and actor Arnold Schwarzenegger. Harris also garnered positive attention for spending Election Night at Howard University with her supporters, in addition to leading in polls in three crucial swing states. On the negative side, President Biden’s comment referring to Trump supporters as “garbage” was among the leading negative stories.
The methodology used by the "U.S. Election Page," developed by independent researchers, combines artificial intelligence techniques with sentiment analysis to identify the purpose of the articles and the biases in favor of one candidate or the other.
