Uncanny Index

The formal instrument proposed in Traversing the Uncanny Ridge for locating novelty between systems. Take a pair of systems' joint output (X, Y) and an observer's prior density p; Bayesian surprise is E[−log p(X,Y)]. The uncanny index Δ((X,Y), p) is the expected change in that surprise relative to a null-protocol counterpart — the same pair run independently — so positive Δ means the interaction made outputs less expected than independence, negative Δ means hyper-convergence. Sup and inf over all admissible pairs give upper and lower indices for a protocol. It makes novelty explicitly perspectival (a function of p) and, by itself, cannot tell novelty from noise — the "ridge" is a peak between collapse and chaos that the formula does not encode.