The Reflexive Integrated Information Unit: A Differentiable Primitive for Artificial Consciousness

TL;DR

The paper introduces the Reflexive Integrated Information Unit (RIIU), a differentiable module for artificial consciousness that enhances information integration and causal footprint recording, demonstrated to improve recovery speed in a Grid-world task.

Contribution

The RIIU is a novel, trainable recurrent cell that incorporates causal footprint tracking and information integration, enabling empirical exploration of consciousness-like computation.

Findings

RIIUs are end-to-end differentiable and composable.

RIIUs outperform GRUs in reward recovery speed after failure.

RIIUs maintain a non-zero Auto-Φ signal indicating integrated information.

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 $h$ with two additional vectors: (i) a meta-state $\mu$ that records the cell's own causal footprint, and (ii) a broadcast buffer $B$ that exposes that footprint to the rest of the network. A sliding-window covariance and a differentiable Auto-$\Phi$ 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 $\Phi$-monotone plasticity under gradient ascent. In an eight-way Grid-world, a four-layer RIIU agent restores $>90\%$ reward within 13 steps after actuator failure, twice as fast as a parameter-matched GRU, while maintaining a non-zero Auto-$\Phi$ signal. By shrinking "consciousness-like" computation down to unit scale, RIIUs turn a philosophical debate into an empirical mathematical problem.