Navigation poses a difficult problem for the human visual system, which must process complex and noisy scenes while keeping pace with a relentless stream of incoming information. Since not all information is equally useful, it must allocate its limited resources selectively, which leads to simplified and distorted spatial representations. A means for estimating the detailed structure of these internal representations in the context of realistic and ecologically valid settings is critical to understanding and predicting what visual information biases people in the wild, yet it also poses two difficult challenges. The first is that traditional experimental approaches for estimating spatial memory distortions typically used highly controlled and simplified visual environments, making it hard to generalize their findings to the real world. The second is that they often relied on simple parametric models fit to limited amounts of data, which are not expressive enough to provide detailed and complete estimates, especially in more complicated settings. Here we use the Unity gaming engine as an ideal environment for extending existing work to more ecologically valid and complex 3D scenes. The solution to the second challenge comes from a novel experimental paradigm that produces rich non-parametric estimates of spatial memory distortions that have eluded traditional approaches and can reveal collective memory distortions in unprecedented detail, even in rich and complex 3D environments. The goal of the proposed projects is to adapt this existing experimental paradigm for uncovering collective visual-spatial and navigational memory distortions in an immersive and ecologically valid environment.

Collaborators

Thomas Langlois (Princeton University), Tom Griffiths (Princeton University)