When Does the Body Bias the World? A Bayesian Causal Inference Model of Embodied Spatial Perception

<p>Bodily states can bias spatial judgments, yet the mechanisms underlying such effects remain contested. Using the backpack effect in hill-slant perception as a motivating case, the present study formalized a multisensory-integration account of embodied spatial perception within a Bayesian causal inference (BCI) framework. The model treats visual information specifying hill slant and bodily information associated with anticipated climbing effort as noisy signals that may arise from either a common cause or two independent causes. Deterministic parameter-space simulations examined how predicted perceptual bias varied with cue discrepancy, visual uncertainty, bodily uncertainty, and the prior probability of a common cause. Bodily bias emerged when the bodily signal received sufficient reliability weight and the two signals were inferred to share a common cause. Bias was suppressed when visual information was highly reliable, bodily information received little effective weight, or the common-cause prior was low. Cue discrepancy and bodily uncertainty produced nonlinear effects. The standard BCI model predicted a transition from attraction toward the bodily signal to near-zero bias under extreme discrepancy, but not repulsion. An exploratory extension showed that repulsion can emerge when causal segregation is supplemented by a conflict-sensitive contrastive-correction process. By translating a qualitative multisensory proposal into a formal computational model, we provide a mathematical basis for formulating and testing hypotheses about embodied spatial perception and, more specifically, explicit computational boundary conditions for bodily bias.</p>