PhD Alum Yifan Wu Earns Honorable Mention in ACM SIGecom Doctoral Dissertation Award

Northwestern Computer Science alum Yifan Wu (PhD ’25) has received an Honorable Mention for the Association for Computing Machinery (ACM) Special Interest Group on Economics and Computation (SIGecom) 2025 Doctoral Dissertation Award. The annual awards recognize outstanding dissertations in the field of economics and computation.

Grounded in theoretical computer science, Wu’s research starts from uncertainty quantification, or how AI systems express uncertainty in their predictions. From there, she explores the broader question: how should AI communicate information so that people can use it reliably and appropriately?

Advised by Professor Jason Hartline, her dissertation, titled “Trustworthy AI: Foundations from Proper Scoring Rules,” tackled an increasingly crucial problem: when can we trust AI to make or inform consequential decision-making?

“Even when models are accurate on average, they can still behave in ways that are misleading or unreliable in practice,” Wu said. “My dissertation helps shift the focus from simply building accurate models to building reliable systems that support good decisions.”

Drawing on a mathematical construction for evaluating probabilistic predictions called proper scoring rules, Wu developed a framework for understanding, evaluating, and designing reliable, trustworthy AI systems.

“Proper scoring rules can serve as a common language for studying trustworthiness in AI, highlighting the continuing value of classical approaches like game theory and statistical decision theory,” Wu said. “These frameworks make it possible to represent complex parts of modern AI systems as rational decision-makers with objectives, which in turn helps us analyze and improve those systems using principled mathematical tools.”

Her work resulted in three key findings. First, she outlined how to optimally incentivize truthful and high-quality predictions. Second, she demonstrated that the canonical way of measuring whether predictions are well-calibrated can be misaligned with the goal of decision-making. To address this, she developed a calibration measure that directly reflects decision payoffs, which guides the design of a predictive algorithm with improved decision-making guarantees. Finally, by applying game theory to analyze how humans make AI-assisted decisions and why they don’t use recommendation information optimally, Wu explained how to improve human-AI collaboration.

Currently a postdoc with the EconCS group at Microsoft Research, New England, Wu is expanding her work from predictive models to the economic effects of AI agents in markets. As AI agents become more capable, she is studying whether markets may shift away from transparent public interactions and toward more private, conversation-based exchanges.

“What happens when AI agents begin to act on behalf of humans, for example, by initiating conversations, gathering information, or negotiating in markets?” Wu said. “This can change not just individual decisions, but the structure of the market itself.”

Wu’s recognition by ACM SIGecom reflects Northwestern Computer Science’s leadership in CS + Economics research, said Samir Khuller, the Peter and Adrienne Barris Chair of Computer Science at Northwestern Engineering.

"I first met Yifan when she was a visiting undergraduate research student in spring 2019. Subsequently, she applied to and was accepted to our PhD program. Fast forward six years and she has established herself as a leader in the space of how to evaluate the accuracy of different AI models,” Khuller said. “Yifan has accomplished groundbreaking work in the area of establishing trustworthiness in AI models by using scoring rules. I really look forward to seeing more of her exciting research."

Hartline, who also co-advised ACM SIGecom 2022 Doctoral Dissertation Award winner, Modibo Camara (PhD ’22), explained that Wu’s thesis leverages mechanism design to develop foundational understanding of AI. A central research area of ACM SIGecom, mechanism design relates to how algorithms can guarantee good outcomes even when their input is provided by self-interested, strategic individuals with preferred outcomes.

“Prior to Yifan’s thesis, mechanism design primarily found applications in the design of online marketplaces like Uber, Airbnb, and online advertising,” said Hartline, professor of computer science at the McCormick School of Engineering and the director of Northwestern’s Online Markets Lab. “Her work is a call to action to the ACM SIGecom community to apply its methods to AI. I am very optimistic that this research direction will lead to considerable impact, especially on crucial questions related to AI safety, alignment, and trustworthiness.

“Yifan has become a truly exceptional researcher, one of the very top PhD students in theoretical computer science at Northwestern. While collaborating with her over the last few years, it has been a struggle to keep up, but also a tremendous privilege to get to accompany her journey in developing the foundations of AI trustworthiness.”