COLLEGE STATION, Texas — Data-driven machine learning can analyze large amounts of information to provide suggestions and help make decisions. However, it can also bring challenges related to bias in the data it uses, potentially leading to discrimination against specific groups.
“The outcome of this piece of research is to enhance the trust of artificial intelligence in broad applications so that other people would like to utilize AI towards the critical or high stake applications,” Assistant Professor Na Zou in the Department of Engineering Technology and Industrial Distribution at Texas A&M University said.
Dr. Zou has been working since 2021 on fair AI techniques. She used to focus on the model-centric technique, where she would create an individual model based on the data given to tackle fairness issues, but now she has changed her approach.
"The difference between data-centric AI and model-centric AI is we will actually pay more attention to the quality of data because that's the source," Zou said. "We want to improve the data quality so that we can remove the bias from the source.”
The importance of bias in AI can’t be understated with how prominent the technology is going to be in our everyday lives in the future. Machine learning can impact sectors of the economy, the healthcare industry, the education system, and even employment opportunities.
“We have limited prior knowledge of the bias in AI," Zou said. "We will also want to work out that and then later we want to incorporate our understanding of data, on prior knowledge, into the entire modeling process to enhance the fairness issues.”
To help support her research, the National Science Foundation awarded Dr. Na Zou with a grant worth $547,741 in late April.
Also on KAGS: