Imagine you are asked to solve a Rubik’s Cube. For most of us, the task is clear: rotate the cube’s faces to make each side a single color.
What happens, however, when someone unfamiliar picks up the iconic puzzle and is asked to make each side one color? Do they shift the sides around like one of us, or do they approach the cube another way? Do they rearrange the colored stickers of the cube? How about painting each side a single color? Do they look for a button that magically changes the color of each face?
When a user’s task is not clearly defined, unintended consequences can occur. That is what Â鶹´«Ă˝ Professor Heng Xu—one of the world’s foremost cybersecurity experts—reveals through her current research on how algorithms respond to tasks.
A professor of information technology and analytics in AU’s Kogod School of Business, Xu also serves as director of the Kogod Cybersecurity Governance Center, overseeing the university’s initiative centered on cybersecurity and privacy research.
Through her work in artificial intelligence (AI) governance, privacy protection, data ethics, and fairness in machine learning, Xu is redefining corporate social responsibility in today’s AI era. Her expertise is so interdisciplinary that it doesn’t come as a surprise Xu was recently introduced at a professional conference as a unicorn. Her work bridges disciplines beyond business, including computer science, law, and psychology.
Professor Xu and a team of collaborators are currently in the middle of a $1 million three-year project examining fairness and bias in AI hiring systems. Funded by the National Science Foundation and Amazon, Kogod is the only business school among the awardees’ lead institutions.
Xu’s team combines their data ethics, machine learning, and human resource management expertise to find solutions valuable to the technology and business communities. By focusing on AI in the business domain of human resource management, their work has the potential to effect positive change in hiring practices worldwide.
As more businesses wish to implement AI to improve efficiency, Xu’s study becomes even more essential to ensure results that are both fair and without unintended consequences.
Just like an unfamiliar person solving a Rubik’s Cube, if the task and objective are not made clear, AI can respond in unexpected ways. “What we are telling the AI to do may differ from what we think or assume we are telling the AI to do,” she noted. Xu is also finding that, like humans, AI has limitations. Take the household vacuum cleaner, for example. A self-operating robotic vacuum cleaner may be excellent at removing dirt from beneath a sofa or bed, but it cannot effectively reach the corners of a room. On the other hand, a traditional vacuum operated by a human can effortlessly reach corners but cannot clean beneath furniture with ease.
These examples align with Xu’s discoveries that are backed by case studies and testing in AI-based hiring systems. What does this mean for real-world applications? Xu noted that technologists must return to the design room and redefine tasks together with domain experts to achieve desired outcomes. AI should also be used in tandem with human resources professionals. For example, AI can improve efficiency in the preliminary screening of resumes by way of speed and accuracy. But a human’s expertise and training remain vital to the final decision-making on whom to hire—and to guaranteeing a fair hiring process.
Xu’s work can help shape national policy in a variety of areas, including AI governance and data protection standards to regulate and assess risk levels, ensure access to information, and maintain the ethical use of data. “Washington, DC, is the place where our research findings can quickly translate into implications for policymakers,” she noted about AU’s hometown access to lawmakers.
She also noted that building bridges between disciplines is both extremely rewarding and frustrating. “You can end up with neither side recognizing your work,” she said. “[Philanthropic support] can really facilitate early stages of exploration, discovery, and bridge-building that have not yet matured enough to be recognized by the scientific peer-review process…and be rewarded by these competitive grants,” stated Xu, referring to her current project’s funding. “This [bridge-building] work is so unique in this space…it needs time to convince the world to buy into it.” Â
Through Change Can’t Wait, the AU community is providing the resources for extraordinary scholar-teachers like Heng Xu to take risks and push their research in novel directions and new areas of exploration. AU faculty guide the next generation of changemakers and help create meaningful change in the world.
To learn more about how you can support AU’s faculty, contact giving@american.edu or make a gift online to the .
