Melissa Pappas
The U.S. AI Safety Institute Consortium (AISIC), initiated on February 8, 2024, unites AI creators and users, academics, government and industry researchers, and civil society organizations in support of the development and deployment of safe and trustworthy artificial intelligence. The consortium is housed under the U.S. AI Safety Institute (USAISI) and will contribute to priority actions outlined in President Biden’s landmark Executive Order, including developing guidelines for red-teaming, capability evaluations, risk management, safety and security, and watermarking AI-generated content.
The Consortium brings together more than 280 organizations to develop science-based and empirically backed guidelines and standards for AI measurement and policy, laying the foundation for AI safety across the world. This will help prepare our country to address the capabilities of the next generation of AI models and systems with appropriate risk management strategies.
The University of Pennsylvania has been invited to join these collaborating institutions in the consortium, and will be represented by Danaë Metaxa, Raj and Neera Singh Term Assistant Professor in Computer and Information Science (CIS). Metaxa brings their expertise in bias in AI-based sociotechnical systems, such as targeted ads and social media platforms, and will contribute their expertise on algorithm auditing, a method they have explored through sociotechnical auditing research. This work values the human user’s experience across platforms that include AI-generated content and is featured in Metaxa’s book, “Auditing Algorithms: Understanding Algorithmic Systems from the Outside In.”
“I am eager to make contributions in several areas the National Institute of Standards and Technology lists, including AI safety guidelines, benchmark datasets, tools for evaluating AI, and evaluation methods,” says Metaxa. “For example, my research group and team of collaborators at Penn has recently conducted studies that address a few of these areas, including using our custom tool to audit real users’ exposure to online media, and two studies that establish methods for auditing GPT-produced content as well as LLM content moderation pipelines.”
Metaxa’s contributions will support specific goals of the consortium, such as engaging in collaborative and interdisciplinary research, identifying evaluation requirements and approaches that allow for a more complete understanding of AI’s impact on society and the economy, and facilitating the transfer of technology and data between consortium members. As a collective group dedicated to knowledge sharing and ethical use of data and AI tools, the work of the consortium will be open and transparent and welcomes interested parties to work together in building and maturing a measurement science for trustworthy and responsible AI.
“AISIC is a bridge between academia, industry and government,” says Metaxa. “It is important we have these bridges to ensure our research translates to real-world impact, and I am honored to bring my expertise to the consortium and represent Penn in this collaborative effort of keeping AI safe and ethical.”