The paper asks a deceptively simple question: if you are standing somewhere inside a two-dimensional region and you repeatedly release tiny simulated particles that wander randomly (a process called Brownian motion) until they hit the boundary, can you figure out where you are standing just by watching how long those particles take to escape? The time a particle spends wandering before it exits is called an exit time, and its statistical distribution carries geometric information about the surrounding space. The authors want to know whether that information is rich enough to pinpoint your location.
The main results show that for a surprisingly broad collection of shapes, the answer is yes, at least up to symmetry. For regions bounded by conic curves (ellipses, parabolas, and hyperbolas), as well as wedge-shaped regions and equilateral triangles, just the first two statistical moments of the exit time (essentially the average exit time and the average of its square) are enough to determine your location. For rectangular regions, you need the full infinite sequence of moments, but that sequence still does the job. The phrase "up to symmetry" means that if the region has a mirror symmetry, you cannot distinguish two points that are mirror images of each other, which is an unavoidable limitation.
The technical engine behind all these results is something the authors call a "Poisson hierarchy." This is a structured family of partial differential equations, each one building on the solution of the previous, that links the statistical moments of exit times to geometric and spatial information. By solving this hierarchy systematically, the authors can extract location from the probabilistic data in a rigorous way. The work connects probability theory, geometry, and the theory of partial differential equations, and it can be seen as a probabilistic cousin of the famous question of whether you can determine the shape of a drum just by listening to it.