Jordan Levine, MIT Lecturer and Partner at Dynamic Ideas, outlines why he believes executives and regulators must do more to combat AI bias – and what they can do about it.
When the EU announced its proposed new AI legislation in April 2021, the bloc touted the new laws as a necessary step to ensure Europeans can trust AI technologies. But for Jordan Levine, Partner at consulting firm Dynamic Ideas, the proposals are something of a ‘blunt instrument’.
In this week’s Business of Data podcast, Levine argues that this kind of legislation is, at best, a starting point. It’s up to AI-focused executives to sit down and implement practical frameworks for ensuring AI is used responsibly in their organizations.
“I'm 100% supportive of the government getting involved in establishing the rules,” he says. “[But] I hope that both academics and business [and] society-conscious individuals get excited and say, ‘OK, how do we refine this?’”
In Levine’s experience, there are many things that can cause ethical issues when enterprises put AI or analytics models into production. That’s why much of the work he does at Dynamic Ideas is geared toward educating people about AI bias challenges.
He says it’s important for businesses to have both clear mitigation strategies to combat ethical issues such as biased decision-making and the right tools or technologies to orchestrate those strategies in practice.
“What I try to do is show how to mitigate those issues and then show actual techniques that exist today, [so] that you can leverage open-source software to do the processing,” he says.
Levine argues that business leaders must use a framework like the one he’s developed to make sure they are aware of the ethical issues that may arise from the ways they’re using AI and analytics. This will allow them to take steps to make sure these issues are addressed.
“I hope they can use this framework to actually challenge their analytics groups,” he says. “To actually sit down with the individuals writing the algorithms and confirming whether the issue does or does not exist.”
However, Levine concedes that no framework for combatting AI bias can ever really be complete. Technology is constantly evolving, and enterprises are constantly innovating with it. So, AI-focused executives must be vigilant and reevaluate their AI practices regularly with an ethics lens.
Levine concludes: “The more precise that we can get in terms of bias and ethics and the more, the more discrete issues we can identify and then think through how to mitigate them and show examples of mitigation, I think, the better we all are.”
Key Takeaways
· Regulatory compliance is not the same as ethical behavior.Enterprises must go beyond what’s required of them by law to ensure their AI practices are ethical
· Executives must be aware of potential ethical issues. If executives don’t know the specific risks that come with adopting AI technologies, they will struggle to ensure the right processes are in place to mitigate them
· AI ethics frameworks must be updated regularly. AI-focused executives must constantly reevaluate their AI ethics strategies to ensure their teams are following current industry best practices