Artificial intelligence (AI) and machine learning pervade almost every aspect of daily life. Unfortunately, recent studies show that the algorithms which drive AI technology can exhibit racial and gender bias, discriminate within a computer-vision facial recognition system, and encode gendered bias in natural language processing.
That is why Drexel College of Computing & Informatics (CCI) Associate Professor Edward Kim, PhD (PI), Assistant Professor Matthew Burlick, PhD (co-PI) and Professor Dario Salvucci, PhD (co-PI) are working to address racial and gender bias in algorithms. Their project, titled “Bias in the Machine,” is the recipient of a Racial Equity Project Award by Drexel University’s Rapid Response Research & Development Fund.
Artificial intelligence is “a sense and react system,” as Kim describes, while machine learning, as a subset of AI, works to learn the parameters of that system. We can find very basic artificially intelligent systems in everyday items like thermostats, which sense and adjust to a designated temperature; through machine learning, the thermostat could be trained with data on the most preferred temperature setting and adjust accordingly.
AI and machine learning technology can have more significant and life-altering ramifications, especially if algorithms exhibit bias. Kim offers the bank loan approval process as an example:
“The bank’s algorithm learns by looking at the data that people of a certain race, gender, zip code, etc., default more frequently on their loan. And now, when I come asking the bank for a loan, I may be unfairly denied because of a bias in the system that really has no bearing on my personal application and shouldn't be taken into account. The goal of our work is to raise awareness of bias in artificial intelligence and machine learning, and to research ways that computer scientists can address and mitigate these concerns.”
Drexel's Rapid Response Research & Development Fund was designated for urgent action, short-term projects. In honor of Juneteenth and with a deep commitment to the needed examination and eradication of racism in this country and around the world, Drexel's Rapid Response Research & Development Fund focused on racial equity projects. The Fund received 33 submissions total, of which 22 were awarded.
“I am excited for this opportunity to contribute toward progress in racial equity,” said Kim. “Often times it is easy to ask the question ‘can we make a machine learning algorithm that performs a certain task?’ and ignore the harder question of, ‘should we being making this in the first place?’ Or, ‘how can we ensure that societal and ethical considerations are taken into account?’ Personally, I'm glad that Drexel University is asking and supporting the hard questions.”