In a recent experiment by a research team at BYU, artificial intelligence You can predict how different demographics will vote in elections.
This research, conducted by a team of BYU political and computer science professors and graduate students, explored ways to use AI as a surrogate for human responders in survey-style research.
ChatGPT 4 displayed on a smartphone with the OpenAI logo on the screen in the background in Brussels, Belgium. (Jonathan Raa/NurPhoto via Getty Images)
To see if this is possible, the team tested the accuracy of the programmed algorithms. GPT-3 modelwhich mimics the relationships between human ideas, attitudes and sociocultural contexts of different demographics.
In one experiment, researchers created artificial personas and assigned attributes such as race, age, ideology, and religion. Next, using data from American National Election Studies (ANES), the team tested whether their “persona” voted in the same way as he did in the 2012, 2016, and 2020 US presidential elections. bottom.
AI could become a ‘terminator’ and could dominate humans in Darwin’s laws of evolution, report warns
Ultimately, the researchers found a “high correspondence” between how AI personas voted and how American citizens voted in those elections.
David Wingate, a computer science professor and co-author of the study, said he was “completely surprised” by how exactly the experiments matched.
It is worth noting that the algorithmic model was not trained to “do political science”, but “hundreds of billions of words of text downloaded from the Internet.”
“[T]The consistent information we got was very relevant to how people actually voted,” he said.
Artificial intelligence words are shown in this picture taken on March 31, 2023. (REUTERS/Dado Ruvic/Illustration)
Another experiment yielded very similar patterns between human and AI responses to interview-style survey questions.
The team holds the promise that the results of their experiments will help researchers, marketers, and pollsters formulate better survey questions or “simulate hard-to-reach populations.” I believe you are.
“We are learning that AI can help us understand people better,” said BYU professor Ethan Busby. It’s about augmenting our abilities rather than replacing them.
“By allowing us to test our surveys and messages in advance, we can work more effectively with people.”
CLICK HERE TO GET THE FOX NEWS APP
A study titled “Out of One, Many: Using Language Models to Simulate Human Samples” was published in a journal political analysis.
