Carnegie Mellon’s Sherry Tongshuang Wu Wins Okawa Research Grant to Advance AI Agents for Real-World Decision-Making

Sherry Tongshuang Wu, an assistant professor at Carnegie Mellon University’s Kenneth C. Griffin School of Computer Science, has been awarded an Okawa Research Grant to investigate how artificial intelligence “agents” can better support real-world decision-making for individuals and groups.

AI agents are increasingly positioned as digital helpers for tasks that involve many options—such as organizing travel itineraries, weighing competing products, or coordinating other multi-step choices. Yet these systems often assume that a user’s goals and preferences are known and stable from the start. In practice, people frequently discover what they value only after seeing possibilities, and their priorities can shift as they learn more about costs, constraints, and trade-offs.

Wu’s project targets that gap: improving how AI agents infer and refine human preferences over time and how they communicate the compromises embedded in complex choices. Rather than simply producing a single recommendation, the work will explore ways for agents to surface meaningful alternatives and help users understand why one option might be better on one dimension but worse on another. The research also emphasizes the interactive nature of preference discovery—how an agent can respond when a person provides feedback or changes their mind.

A central deliverable of the grant-funded work will be an open-source evaluation framework designed to measure these capabilities systematically. The framework will employ simulated users—software models that stand in for people with evolving and sometimes uncertain priorities—to test how well AI agents can (1) elicit what matters to a user, (2) present and compare options, and (3) adapt when new information or feedback is introduced. By making the tools open source, Wu aims to support broader progress and reproducibility across the field.

The research will also extend beyond one-on-one assistance to team decision-making, where challenges multiply. Groups may need to reach consensus despite members placing different weight on competing goals. Wu’s work will examine how AI agents might help teams navigate these situations, supporting deliberation when priorities clash and decision stakes are high.

Wu is part of CMU’s Human-Computer Interaction Institute, where her wider research agenda focuses on human-AI collaboration—especially the practical question of how people can work effectively with AI systems that are useful but imperfect. Her stated aim is to develop tools and evaluation methods that keep humans in control of the process while enabling better-informed decisions.

The Okawa Research Grant is administered by the Okawa Foundation for Information and Telecommunications, which funds research in information and communications technology. Additional details about the program are available through the Okawa Foundation.

Why it matters

As AI agents move into everyday planning and workplace coordination, the biggest risk is not just wrong answers—it’s systems that nudge decisions without accurately capturing what people actually value. Tools that evaluate preference discovery and trade-off explanation could set higher standards for transparency and user control. Watch for whether open benchmarks from this work become widely adopted for testing next-generation agentic AI.