ľ¹Ï¸£ÀûÓ°ÊÓ

PLEASE NOTE: If a candidate gives a layman's talk, the livestream will start fifteen minutes earlier.

Artificial intelligence (AI) and machine learning are increasingly part of our daily lives, influencing everything from medical diagnoses to self-driving cars. However, understanding how these systems make decisions is often challenging because their inner workings remain hidden.

Decision maps are visualization techniques that show AI decisions by turning complex processes into simple, colorful maps that illustrate how AI distinguishes between options—for example, deciding whether an image shows a pedestrian or an obstacle in autonomous driving. My research showcases the applications of decision maps in a geology research context, evaluates these decision maps, and uncovers a significant limitation: existing maps only cover a restricted portion of the data space, potentially leading to incomplete insights.

To address this issue, I developed a novel controllable approach allowing users to explore and understand these AI decisions more comprehensively. This enhanced method not only breaks the abovementioned limitation but also enables broader applications. By making decision maps more versatile and informative, my work helps everyone better understand and trust AI systems.