eyeflick dizzyThe Year 5 student at Kinsale Community School combined her interests in politics and computer science to create the project. She set out to use artificial intelligence to address the growing problem of bias in news media, resulting in an Oracle Academy Special Award and third place in the Senior Technology Individual category. “Bias in the news is a bigger problem today than ever before, and I wanted to use my interest in computer science to address this issue,” Eiflick said. Carrigdoon.
“Even though we think we can recognize bias when we see it, humans are actually very bad at identifying it ourselves.”
During her research, Eiflick found that just over a quarter of adults can accurately identify bias in statements presented as factual information. As a result, many people consume biased news without realizing it, and misinformation spreads unnoticed. This motivated her to develop an AI tool that can alert readers to possible biases in the news they read.
iFlick wrote all the code for the project himself and trained the AI model independently. She learned machine learning and natural language processing through online courses and resources. She didn’t have access to a GPU at home, so she trained her model using free GPU services available online.
“This made the project even more difficult, but it was the best way to approach it,” she said. “It forced me to learn a lot about AI and machine learning, skills that I might not have learned otherwise. AI is becoming an integral part of computer science, and I will be grateful in the future that this project helped me develop those skills.”
As part of his findings, Eiflick found that models built using a new architecture focused on analyzing language characteristics outperformed existing state-of-the-art solutions. It also showed better generalizability. In other words, it performed well even with content that was different from what it was trained on.
“I was pleasantly surprised to see that test performance also improved, as increasing versatility was the main goal of my project,” she explained. “These two measures often contradict each other, so it was interesting to see them both improve at the same time. My results highlight the potential of this linguistic approach for bias identification and other tasks in computational linguistics.”
Biased news is a widespread problem that affects everyone who consumes media.
“My project will provide a useful tool to help people recognize implicit bias in the news they read, and help reduce the spread of misinformation,” Eiflick said.
“The approach I used is very novel and has great potential for future applications in computational linguistics.”
