Uncanny Ridge

Coined in the Cognitive Infrastructures studio as the inverse of Mori's uncanny valley. Where the valley is a trough of discomfort, the ridge is "a peak of interactional intensity" — the zone where intentional misrecognition between heterogeneous systems (models, agents, humans) reorganizes intelligibility rather than breaking it, producing perspectival novelty. Its foil is the canny valley: hyper-convergence between user expectation and model output, the aesthetic collapse caused by RLHF mode collapse and CLIP guidance. The paper proposes an uncanny index — the expected change in Bayesian surprise under an interaction protocol versus a null-protocol counterfactual — to locate the ridge empirically. Distinct from the journal's inverse uncanny valley, which is about seeing oneself estranged through machine eyes.