Humans Navigating Complexity with AI Helpers

Modern particle accelerator control rooms are like busy air traffic control centers, where skilled operators juggle hundreds of interconnected and time-sensitive tasks to generate and shape high-energy particle beams. Control room tasks are increasingly performed with the support of AI. This talk explores our study of how operators select and organize these tasks, revealing key differences in task understanding based on expertise. Our study suggests that novices stick to “safe” task groups, while experts have more precise mental models of the entire complex system, allowing them to perform “riskier” tasks to achieve better results. Insights into how humans understand and think about the complex problem of accelerator operations and how these mental models evolve with expertise can help guide the design of trustworthy AI helpers that complement human expertise rather than replace it. Through this human-centered AI approach, I will share how we can make high-stakes technological environments safer and more efficient.

Speaker: Wan-Lin Hu, SLAC

Register to attend in person, or see weblink to watch online