Hybrid AI with Biochemical Modulation: The Impact of Hormonal Analogues on Attention and Decision-Making in Transformer Networks

Abstract

This study presents a practical implementation of an innovative hybrid AI architecture integrating traditional transformer neural networks with biochemical modulation inspired by hormonal processes observed in plants. By simulating emotional and cognitive states using analogues of dopamine, serotonin, and cortisol, we analyzed their impact on the network's attention mechanism and predictive abilities. Results demonstrated that biochemical signals significantly affect network behavior, closely mimicking biological systems. This work extends previous theoretical frameworks regarding hybrid AI architectures (Seidl, 2025a) and a functional definition of consciousness (Seidl, 2025b).