What is artificial intelligence? Is it an innovation that will lead to a robot populated utopia, is it just another technological tool made up of ones and zeros or something in between?
Professor Meredith Broussard was among those who tackled this question at the Jordan Center for Journalism Advocacy and Innovation’s symposium “Addressing the Impact of Social Media and Artificial Intelligence on Democracy” at the University of Mississippi on April 2-3.
Broussard is a data journalist, associate professor at the Carter Journalism Institute in New York and the author of several books including “More than a Glitch: Confronting Race, Gender and Ability Bias in Tech.”
Broussard gave her definition of AI.
“We get very worked up over imaginary concepts of AI,” Broussard said. “We get confused with Hollywood concepts of AI, but really all an AI system is doing is a lot of computational statistics.”
Broussard’s work often focuses on analyzing the data sets that AI systems train on to do their calculations. Often, these data sets are biased, Broussard said.
“We don’t live in a perfect world, so the data that we’re using to train AI systems is data from our imperfect world,” Broussard said. “Then the AI perpetuates these biases.”
Broussard was featured in the Emmy-nominated documentary “Coded Bias,” which details how commercial facial recognition AIs from companies like IBM have failed to accurately identify people of color. These biases have real world impacts, especially when the systems are used for surveillance and policing. Broussard talked about the complexity of addressing AI bias and its real world impacts.
“IBM, to their great credit, said, ‘We are going to increase the diversity of our training data.’ Lo and behold, the AI got more accurate.” Broussard said. “On the other hand, we don’t necessarily want the AI getting more accurate. If something like facial recognition is used in policing, then it’s going to disproportionately affect people of color negatively. Maybe a better solution is not to use facial recognition in policing at all.”
Broussard also talked about the AI systems used for mortgage approvals. In this case, the data to make AI systems more equitable isn’t available, Broussard said.
“Mortgage approval algorithms tend to deny borrowers of color at higher rate. As a data scientist, you might hear this and think the problem is in the training data,” Broussard said. “The problem is that we don’t have better training data where there hasn’t been financial discrimination in lending.”
However, Broussard did note some projects that are diversifying their data sets to the benefit of users.
“Languages that are not spoken by very many people are generally not included in AI systems,” Broussard said. “There are a number of projects around creating data sets for languages that not many people speak.”
Broussard also talked about how journalists can use AI to analyze large data sets simultaneously, help make transcripts and translate other languages.
While there are effective uses of AI, Broussard warned people of getting wrapped up in what she calls the “AI hype cycle.”
“People are talking a big talk about how transformative (AI) is going to be, but when you hear people making these enormous claims, you really need to be a little skeptical,” Broussard said. “When it comes to technology, people are going to over promise and under deliver.”
While many imagine that AI will bring more jobs and a “sleek, technologically enabled future,” Broussard said AI has reduced the amount of jobs available. A 2023 report from the Pew Research Center found that 19% of American workers were in jobs that are the most exposed to AI.
Broussard said that job loss will lead to a reduction in something that we all need — human interaction.
“One of the things that makes us happy is interacting with other human beings,” Broussard said. “This technological future where machines do everything and people sit home by themselves … that’s not a good way for human beings to live.”
