The Reflexive Integrated Information Unit: A Differentiable Primitive for Artificial Consciousness
Abstract
Research on artificial consciousness lacks the equivalent of the perceptron: a small, trainable module that can be copied, benchmarked, and iteratively improved. We introduce the Reflexive Integrated Information Unit (RIIU), a recurrent cell that augments its hidden state with two additional vectors: (i) a meta-state that records the cell's own causal footprint, and (ii) a broadcast buffer that exposes that footprint to the rest of the network. A sliding-window covariance and a differentiable Auto- surrogate let each RIIU maximize local information integration online. We prove that RIIUs (1) are end-to-end differentiable, (2) compose additively, and (3) perform -monotone plasticity under gradient ascent. In an eight-way Grid-world, a four-layer RIIU agent restores reward within 13 steps after actuator failure, twice as fast as a parameter-matched GRU, while maintaining a non-zero Auto- signal. By shrinking "consciousness-like" computation down to unit scale, RIIUs turn a philosophical debate into an empirical mathematical problem.
Cite
@article{arxiv.2506.13825,
title = {The Reflexive Integrated Information Unit: A Differentiable Primitive for Artificial Consciousness},
author = {Gnankan Landry Regis N'guessan and Issa Karambal},
journal= {arXiv preprint arXiv:2506.13825},
year = {2025}
}