Confronting and combatting algorithm bias at CoNECD conference

Body

Algorithms help us make hard decisions every day. Credit card companies, job boards, and more use fast-thinking algorithms to fairly decipher who fits their chosen criteria. But sometimes, they aren’t always as fair as they appear. 

At the annual Collaborative Network for Engineering and Computing Diversity (CoNECD) conference in January, the Director of the MS Data Analytics Engineering program, James Baldo, presented how algorithm biases arise and where data analysts and algorithm creators could make changes to increase fairness. 

“I looked at algorithm bias from a high-level technical perspective to show the audience that yes, algorithms can be biased, but that there is more to it than a yes or no analysis,” says Baldo. 

There are numerous aspects of algorithms that could hold bias. The data used for the algorithm, the core of the algorithm itself, and even the people interpreting the algorithm’s data could be where bias sneaks into the decision-making process, says Baldo. 

“Algorithms use artificial intelligence and are designed by computational data that may have an inherent and unintentional bias,” he says. “Employment decisions are a good example. If an algorithm is using a sample of data to sift through applicants for a software engineering job that doesn’t include many women, the algorithm could unintentionally sift women out.” 

Baldo says there have been numerous studies on algorithm bias. Still, he shared his thoughts at the conference because he felt it was important for conference attendees to understand algorithm bias’s root causes.  

“One slide I presented discussed how we achieve fairness with algorithms. We need to look at the data and try and detect biases in it. We can train people who interpret the data and educate them on the best practices,” says Baldo. “I mainly wanted my presentation to raise awareness, and I wanted to take some of the mystery out of algorithms.” 

From a data perspective, Baldo works to embed these best practices into the data analytics engineering program. “We are trying to embed education on the effects of algorithm bias into the MS program. Since it is an interdisciplinary program, we have to work together to figure out how to do that collaboratively.” 

Baldo sees algorithm bias awareness and prevention as crucial for building the data analytics workforce. “The social fabric of this is very important, and we have a responsibility as engineers to address it, and this was a start.”