How ChatGPT served up false information before the 2024 election
On Oct. 25, 2024, 10 days before the election that would return Republican Donald Trump to the White House and more than three months after Kamala Harris ascended to the top of the Democratic ticket, ChatGPT was still telling users that Joe Biden was running for a second term as president.
In the weeks leading up to the 2024 U.S. presidential election, ChatGPT, a generative AI model, was found to be providing inaccurate information, raising concerns about its role in a democratic society. A study co-authored by University of Massachusetts Amherst journalism scholar Heesoo Jang and researchers from the University of North Carolina at Chapel Hill examined ChatGPT's responses to election-related prompts.
They discovered that 9.6% of 592 verifiable claims made by the AI on October 25, 2024, were incorrect. The model inaccurately stated that Joe Biden was still in the race for a second term, claimed that North Carolina Governor Roy Cooper was seeking reelection (despite him being term-limited), and omitted critical information about efforts to overturn the 2020 election.
ChatGPT also used hedging language, such as "could potentially" and "generally aligned," and in some instances, omitted key details to obscure political realities. Furthermore, the model struggled to recognize redefined and politicized terms, like the "DEI hire" label used against Kamala Harris. The researchers argue that while ChatGPT may excel in academic settings, it was not designed to uphold democratic institutions.
They emphasize the need for journalism that not only delivers truth but also helps people evaluate the truth, especially as AI increasingly replaces traditional search engines and links to primary sources.
Written by urgent.news from Phys.org's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
