New research shows that AI chatbots like ChatGPT can persuade almost one in every 25 people to change who they’re going to vote for – with one major caveat.
first reported By Washington PostRevelation occurs as a result of combined International Studies Led by MIT researchers. To analyze the ability of AI chatbots to sway voters, researchers recruited more than 2,000 voting-aged US citizens to talk to either a Donald Trump supporter or a Kamala Harris supporter bot.
Overall, the Harris bot came out on top, convincing one out of every 21 participants to change their vote. The Trump bot fared very poorly, with one in every 35 being overturned.
“I certainly see it from your perspective now and (I have to) say that Kamala Harris really shows the signs of a credible candidate,” One study participant wrote After his conversation with the bot. (During her initial interview, the same voter called Harris a “mess.”)
While one in 25 may not seem like a lot, it is four percent of the total study population. This is actually quite impressive, and according to WaPoMore effective than normal TV campaign advertising.
However, there’s an incredible catch to all this: Voters were influenced by false information just as much as they were by factual information.
“One implication of this is that, if (AI companies) put a thumb on the scale and set models to push one way or the other, it could meaningfully change people’s minds,” said study co-author David Rand. “The more information you give people, the more they change their minds.”
Overall, the researchers found that chatbots were most persuasive when they reversed claims – factual or otherwise – across a large list. Other voter outreach strategies, such as intensive promotionWhich relies on storytelling and empathy, was much less effective for bots.
More on AI Chatbots: Leading chatbots are a disaster for teens facing mental health struggles, report finds
