Paper · In revision
A unifying field theory of attention
The Magnetic Field Model of Attention (MFA) proposes that attentional intensity decays with the inverse square of psychological distance. The exponent α = 2 is derived from three independent principles: geometric flux conservation, neural wiring-cost optimization, and Amari-type field equations. Six field equations unify five previously isolated theories of attention (spotlight, gradient, zoom-lens, load, and resource models) and yield eight falsifiable predictions. A training-free saliency model achieves AUC = 0.793 on the MIT/Tübingen Benchmark.
The point of the theory is parsimony: instead of one model per phenomenon, one law that the phenomena fall out of. Because the predictions are falsifiable and the saliency implementation is measured on a public benchmark, the framework is testable rather than merely suggestive.