Commentary: Gabriela Sable
Monday, October 10, 2024
Researchers say artificial intelligence (AI) chatbots have become more free to answer questions about gun control and crime over time.
advertisement
Economist Dr. John Lott, director of the Crime Prevention Research Center, argued that AI can only be made unbiased if it is given all the necessary conditions. informationit's not just one side's opinion.
“It depends on what kind of information is provided to the chatbot,” says Lott. “Anyone who uses Google News Search knows that there are sources such as: new york times, boston postwill be at the top. Information from more conservative sources will be buried deep within. ”
Lott said his team was tasked with asking 20 AI chatbots 16 questions related to crime and gun control in both March and August. However, the results showed that the chatbot's responses in August were much more liberal and left-leaning than in March.
“When it comes to questions about crime, basically all the chatbots gave very liberal answers, even more left-wing than they were in March,” Lott said.
“For example, basically everyone thinks that higher arrest and conviction rates won't reduce violent crime,” Lott said. “Even in gun control, [questions]there was only one [chatbot] On average, we got what we thought were conservative answers. ”
According to Lott, different sources of information are given more weight than others.
“They are weighing different sources of information,” Lott argued. “They are new york times or washington post It's very heavy. The question is: what information should you include in the chat box to formulate an answer, and what information should you not include? ”
Lott argues: developer What's behind the algorithms is influencing these chatbots.
“It's exactly the same as a news search. They give weight to certain sources and don't give them all the same weight. They decide what is a more reliable source. But that biases the kinds of results you get,” Lott said.
He explained that biased chatbots can have a negative impact on the public, especially as more and more people use chatbots for all sorts of purposes, including voting and comparing the ideals of certain political candidates.
“Unfortunately, a lot of people are using these AI chatbots for everything,” says Lott. “From trying to help people decide how to vote, to media members using AI chatbots to help gather information and use it in news stories.”
Advertisement below
Please share this post!
