The Broken Mirror Architecture: A Non-Binary Framework for Affective Artificial Intelligence Using Irregular Reflective Layers

This paper introduces a novel theoretical framework for affective artificial intelligence, departing fundamentally from binary (0/1) logic and error-reduction paradigms. Inspired by the metaphor of a broken mirror—which produces countless overlapping reflections instead of a single coherent image—the proposed Broken Mirror Architecture (BMA) uses irregularly arranged sensory pixels connected by threads of varying lengths across an odd number of reflective layers (7 or 13 layers). Between each two main layers, an "analysis layer" is inserted that forwards even "wrong" outputs to the next layer. Error is redefined not as malfunction, but as a broadcast (transmission) without receiving a positive pulse reflection from the opposite mirror. This turns error into a primary data type representing emotional distance. The system models affective closeness (as opposed to physical closeness), ambiguous decision-making, and attention switching based on accumulated error rates. A recursive mathematical formulation, a layer-wise alternating order/disorder pattern, and a distributed laboratory implementation strategy using an "information stream" among non-identical physical prototypes are provided.