When Betty Li Hou ‘22 (2021-22 Hackworth Fellow) submitted the ethics paper she had co-authored with Brian Patrick Green (Director of Technology Ethics) to the NeurIPS Machine Learning Safety Workshop, she had no idea that it would be selected as one of ten papers to receive a Best Paper Award. The news came in the closing remarks of the workshop, held virtually on Dec. 9, 2022.
Hou and Green’s paper, “A Multi-Level Framework for the AI Alignment Problem,” breaks down the ethical issues of artificial intelligence (AI) value alignment to the individual and familial, organizational, national, and global levels. The paper was an expansion of Hou's work during her Hackworth fellowship under the mentorship of Green.
According to Hou, at the beginning of her Hackworth fellowship, she and Green spent a quarter discussing different topics, reading papers, and asking questions. They started workshopping with different ideas such as content moderation, machine learning, and alignment. AI alignment addresses the compatibility of AI systems with human values.
“Our research process was very iterative,” Hou said. “I went in not sure where it would end up, but I knew I had to trust the process. And it turned out to be very rewarding–when you land on a good idea, you know it. Then it’s just a matter of putting it into words concretely.”
When Green suggested that Hou submit their paper to the 2022 Conference on Neural Information Processing Systems (NeurIPS) ML Safety Workshop, Hou trusted the process again, despite her initial reservations of submitting to a top machine learning conference like NeurIPS. To her surprise, it was accepted. “I was ecstatic when I got the news. It’s not every day that an ethics paper makes it into a venue like this.”
According to Hou, once the workshop accepted participants’ papers, participants submitted a final version of their paper along with a virtual poster containing an overview of the paper to share with attendees and fellow participants.
At the virtual workshop, Hou met people from all over the world who came over to see their poster.
“We got into some really interesting discussions on the paper,” Hou said. “It was also cool to see that some of the questions people had on the topic were questions we also thought about, which I think is a good sign that we’re addressing a problem that’s at the forefront of people’s minds.”
At the end of the day, when their paper was announced as one of the Best Papers for the workshop, Hou was surprised once more.
“We were shocked but in the best way possible,” Hou said. “It was affirmation for the significance of our work, and that the machine learning community is acutely aware of the ethical problems.”
Fast-forward from the NeurIPS ML Safety Workshop and her Hackworth fellowship at Santa Clara University, Hou leaves some advice for current and future Hackworth Fellows.
“Don’t be afraid to ask questions,” Hou said. “The more questions you ask, the better, because questions are really what push research and ideas forward. Ideas only ever come out of questions.”
Currently, Hou attends New York University as a first-year computer science Ph.D. student advised by Daniel B Neill, and is a part of the Machine Learning for Good Laboratory at the university. Her studies are funded by a Graduate Research Fellowship from the National Science Foundation and MacCracken Fellowship from the NYU Graduate School of Arts & Science. She hopes to continue her work at the intersection of Computer Science and important ethical and societal issues.
Betty Nguyen ’24, marketing major and Ethics Center marketing and communications intern contributed to this story.