While data science experts create algorithms, embed programs and study machine learning, Duncan Purves asks what ethical issues could arise.
Common technologies use algorithms that are specialized to the intended user. Purves, an assistant professor in the Philosophy Department at UF, picks apart different algorithms and their potential problematic behavior.
“A concern that some people have raised about these systems is that they’re in some ways opaque, or inscrutable, to the people being subjected to them or even the people that are using them,” Purves said.
“We’re interested in a sort of broad concern about fairness and understanding whether and when it applies to these systems.”
Ethical policing
The way computer systems learn patterns from humans can dictate important outcomes.
In one of Purves’s projects, titled “Artificial Intelligence (AI) and Predictive Policing: an Ethical Analysis” evaluates the concern that law enforcement in the United States will use algorithms to incriminate people of color more often than white people.
“The kinds of systems we’re particularly interested in are those that are being used to identify high risk places in jurisdiction,” Purves explained, adding that police officers could use technology as a harassment tactic on offenders. “Predictive policing encompasses systems that are designed to identify high-risk offenders. There are concerns that these systems oftentimes disproportionately target minority communities for additional police attention — this type of targeting is the result of a kind of racial bias.”
Although the project focuses on use of AI and machine learning algorithms in the criminal justice system, Purves studies ethical issues with AI in social media and more. In his upcoming seminar Ethical Issues Facing Data Science, Purves will broadly discuss concerns in AI which researchers from across many disciplines can work together to address.
Collaboration
With UF’s ongoing AI Initiative, Purves said that UF scientists are encouraged to work together to combine different perspectives when studying algorithms.
“One thing that’s key in doing ethically responsible research are these cross-disciplinary collaborations,” Purves said. “People from some disciplines will identify ethical issues, concerns or possibilities that you just wouldn’t recognize from the standpoint of your own discipline.”
This story originally appeared on UF Informatics.
Check out more stories on the UF A.I. Initiative.