Efforts to ensure that artificial intelligence is deployed responsibly and ethically are coming primarily from academic researchers and legislators. That is about to change.
The newly formed Center for Responsible Artificial Intelligence and Governance (CRAIG), described by Northeastern University associate professor of philosophy and CRAIG member John Bustle as the first National Science Foundation-funded research effort of its kind, will combine academic rigor and real-world industry expertise to solve some of AI’s most pressing challenges, experts involved in the research said. CRAIG is tackling everything from technical privacy issues to regulatory issues like never before.
“Companies don’t really have the infrastructure for that,” said Basl, who represents one of the four partner universities leading CRAIG. “What companies have the infrastructure for is the compliance part of complying with existing laws. So the idea was to create a center that would solve industry challenges while also bringing in academia to be part of the solution. … This would be a call to arms to accomplish that.”
In addition to Northeastern University, faculty from Ohio State University, Baylor University, and Rutgers University form the core of CRAIG’s research department. Meta, Nationwide, Honda Research, CISCO, Worthington Steel, and Bread Financial are already involved on the industry side, with more partners joining the center.
Kansu Kanka, director of Northeastern’s responsible AI practice, said responsible AI is typically the first element removed from a company’s AI-related projects. CRAIG addresses this core tension by enabling private industry partners to identify the issues they face in this area. CRAIG researchers propose research projects that address those specific challenges.
“With this structure, we can confidently claim that our research is actually conducted with the same academic rigor that we follow in academia, without any doubt that industry will influence our research or our objectivity,” Kanka said.
One of the many challenges CRAIG researchers aim to address when it comes to responsible AI is homogenization. When a single AI model makes all decisions in a particular industry or field, there is a risk of bias or excluding certain people from consideration.
“Recruiters may not like people who wear purple shoes, but other recruiters might not either,” Bastle says. “But if it’s all the same recruiter, the one with the purple shoes is out.”
Through CRAIG, Basl aims to find ways to measure and mitigate this type of narrow-view decision-making, which can have real impact on AI applications in finance and insurance.
It sounds like a simple model, but Basl and Canca say no model like this exists that combines academic expertise with industry influence. Connecting academic research to real-world applications is always a challenge, especially in emerging fields like responsible AI.
When companies turn to researchers, they are typically in the early stages of responsible AI implementation and are looking for more fundamental solutions. According to Canca, that can be difficult for academics who are familiar with the field and are researching more experimental AI applications.
“Having a center, structure, and authority focused on responsible AI allows us to do novel work, connect this novel work to real industry applications, and get feedback from the applications, which is great,” Canca said.
According to Basl, CRAIG is a win-win for academia and the AI industry. This collaboration will not only help solve important questions about responsible AI; It will create the next generation of workers specifically trained to tackle those very problems. CRAIG plans to support 30 PhD graduates over the next five years. Researchers, Northeastern Co-op, and hundreds more students will participate through summer programs and coursework.
Canca explained that the opportunity to leverage that talent and expertise with an industry leader like Meta who is making such a huge impact in technology is the real promise of CRAIG.
“The dream is to really grow this and really create a broader industry group and a broader community of researchers so that we can set standards, build new tools, and agree on which tools are the best way to use them for which problems, rather than always adopting an experimental style in responsible AI,” Kanka said.
