AI-SDM is one of seven AI institutes awarded today by the National Science Foundation(opens in new window). A five-year, $20 million commitment from the NSF will support the institute.
“The best applications of artificial intelligence in societal domains will come when we not only advance AI for decision-making, but also better understand human decision-making, and when we can bring the two together,” said Aarti Singh(opens in new window), a professor in the Machine Learning Department(opens in new window) of CMU’s School of Computer Science(opens in new window), who will serve as the institute’s director. “Social scientists are studying human behavior. Machine learning researchers are developing new AI technologies to aid decision-making. For maximal impact of these technologies, we need to have social scientists and AI researchers collaborate to come up with solutions that will leverage AI capability while ensuring social acceptance.”
AI-SDM will bring together experts from both the School of Computer Science and Dietrich College of Humanities and Social Sciences(opens in new window) at CMU, as well as Harvard University, Boston Children’s Hospital, Howard University, Penn State, Texas A&M University, the University of Washington, the MITRE Corporation, Navajo Technical University and Winchester Thurston School. This diverse group of researchers and practitioners will work with public health departments, emergency management agencies, nonprofits, companies, hospitals and health clinics to enhance decision-making.
“With artificial intelligence advancing at a dizzying pace, our future depends on researchers, social scientists, decision makers and the public working together to understand these tools and put them to ethical use,” said Congresswoman Summer Lee (PA-12), whose district includes CMU. “I’m proud to announce a $20 million research award from the National Science Foundation for CMU to lead the country’s AI Institute for Societal Decision Making. A collaboration of several institutions, including my alma mater, Howard University, the institute will work interdisciplinarily to design ethical, human-centric AI tools to help improve disaster response and aid public health officials, community workers and clinics.”
AI-SDM is the fifth NSF-funded AI institute to include researchers from CMU, and the first to be led by the university’s expertise. CMU faculty already contribute to the AI Institute for Collaborative Assistance and Responsive Interaction for Networked Groups (AI-CARING), the AI Institute for Future Edge Networks and Distributed Intelligence (AI-EDGE), the USDA-NIFA AI Institute for Resilient Agriculture (AIIRA) and the Institute for Agricultural AI for Transforming Workforce and Decision Support (AgAID). These institutes(opens in new window) were established in 2021.
“The National AI Research Institutes are a critical component of our nation’s AI innovation, infrastructure, technology, education and partnerships ecosystem,” said NSF Director Sethuraman Panchanathan. “These institutes are driving discoveries that will ensure our country is at the forefront of the global AI revolution.”
By bringing together AI and social science researchers, AI-SDM will enable data-driven, robust, resource-efficient decisions and improve outcomes by accounting for human factors that are key to acceptance of these decisions in the field, such as biases, perception of risk, trust and equity. AI-SDM aims to leverage AI to better understand human decision-making; to improve the ability of AI to make decisions; and to apply those advances to create better, more trusted choices.
“Artificial intelligence holds extraordinary potential, and at this critical stage in its development, stakeholders from across disciplinary boundaries must come together to responsibly apply these generational breakthroughs to the real world,” said Theresa Mayer(opens in new window), CMU’s vice president for research. “CMU is grateful for the partnership of the National Science Foundation, whose commitment will allow AI-SDM and its partners to advance powerful, human-centric AI solutions for challenging situations that require split-second decision making.”
Cleotilde Gonzalez
AI-SDM will undertake several foundational thrusts to improve the understanding of human decision-making and create AI tools to assist with it. Cognitive and behavioral scientists will develop computational models to accurately represent how and why humans make the decisions they do in times of crisis. Predicting human choices is key to developing better AI tools and ensuring their success in society. This work will be led by Cleotilde Gonzalez(opens in new window), a research professor in CMU’s Department of Social and Decision Sciences(opens in new window), and Christopher Dancy, an associate professor in the Penn State College of Engineering.
“Our work at the AI-SDM will contribute the foundational research required to accurately predict human choices under conditions of uncertainty, time constraints and temporal dynamics. We will construct the future of experimental and computational cognitive decision science, promoting equity and fairness through human-AI complementarity,” said Gonzalez, who will serve as the institute’s research co-director.
Armed with this understanding, social scientists and AI researchers will work together to understand human-AI complementarity and create models of group and hybrid human-AI decision-making. This will also generate an understanding of how social values such as equity, ethics and risk influence individual and group choices. Leading this work will be Ariel Procaccia, a professor of computer science at Harvard, and Aaditya Ramdas(opens in new window), an assistant professor in CMU’s Department of Statistics & Data Science(opens in new window) and Machine Learning Department.
“When AI or humans predict how a particular situation will evolve or propose varying options to take because of different underlying perceptions of risk and utility, it is important to think about how best to elicit these complex preferences and combine them into a group decision,” Ramdas said. “In a setting where these agents make repeated decisions, we hope to design algorithms that can learn from experience how to combine these decisions — from AI or humans with possibly different individual incentives — toward a common group goal.”
Finally, AI researchers in the institute will develop tools capable of making autonomous decisions that will support people in both disaster and public health management. They will have to work in dynamic and uncertain environments and under intense pressure and constraints. They will have to juggle competing objectives with incomplete information and coordinate with many people using imperfect communication, which is a mighty task for current AI. This work will be led by Sham Kakade, a computer science and statistics professor at Harvard, and Jeff Schneider(opens in new window), a research professor in CMU’s Robotics Institute(opens in new window).
“We are especially excited by the opportunity for AI to help real people in scenarios where good decision-making is most needed, yet difficult to come by,” Schneider said. “Researchers at CMU are already developing advanced AI and robotic systems that will be useful in assisting with societal challenges.”
The AI tools created by AI-SDM will not only assist decision-makers with tasks at hand but will also help them reflect on past actions and evaluate decisions not taken. If an emergency manager or public health official sent resources or targeted interventions at one location instead of another, would the result have been different? Tools that can model or simulate these scenarios will help make better decisions. Extensive research in the humanities has looked at how counterfactual and causal reasoning affect human decision-making and acceptance, and applying this research is key to explainable AI that can be trusted. Kun Zhang(opens in new window), an associate professor in CMU’s Department of Philosophy(opens in new window), will lead this effort.
“Two decades ago, CMU helped create the modern field of causal discovery,” Zhang said. “We are now going to a higher level to find hidden causal variables and causal relations for causal inference and counterfactual reasoning from video, images, text and tabular data. This effort will have direct implications not only in decision-making but also in related fields such as scientific discovery, health care, marketing and more.”