Recent research indicates that the impact of fake news on elections intensifies when disseminated steadily over time without interruption, a study using generative AI models reveals.
While there's no conclusive evidence that disinformation has directly influenced election outcomes, concerns persist regarding its significant influence. With the advent of AI-generated convincing fake videos and efficient dissemination methods, the potential for fake news to sway elections looms large.
Traditional scientific methods face challenges in gauging the extent of this threat due to the inability to conduct repeated experiments in social sciences. However, generative models offer a solution by creating numerous hypothetical scenarios, enabling statistical analysis of different situations.
By modeling voter information access and its transformation into shifts in opinion polls, researchers simulate various election outcomes. Incorporating disinformation into these models unveils its potential impact, providing insights into the likelihood of election result alterations.
One notable finding suggests that while a single release of disinformation might have minimal effect on an election, persistent dissemination significantly influences outcomes. The study indicates that repetitive dissemination of biased disinformation gradually tilts opinion polls in favor of the targeted candidate.
Interestingly, the research suggests that public awareness of the frequency and bias of disinformation can mitigate its impact significantly. Armed with this knowledge, voters become less susceptible to its effects.
While generative models alone don't offer solutions to counter disinformation, they illuminate its potential consequences. Coupled with fact-checking initiatives providing statistical insights into identified disinformation, public resilience against fake news could be substantially bolstered.
More: https://phys.org/news/2024-04-generative-ai-fake-news-impact.html
