Machine Consciousness Matrix: AI Cognitive Theory Benchmarks

Architecture Presets

Theoretical Levers

Recurrent Feedback Depth 0

Metacognitive Probe Resolution 0.40

Predictive Error Minimization 0.15

Causal Integration Density 0.40

Global Workspace Broadcast

Sensorimotor Active Embodiment

Feedforward Pulse

Recurrent Loop

Metacognitive Probe

Active Inference

Functional Reportability vs Causal Integration: Modern autoregressive transformers score high on Global Workspace broadcasting metrics because attention mechanisms mimic a unified cognitive workspace. However, under Integrated Information Theory (IIT), their feedforward computation contains zero causal feedback loops, yielding near-zero intrinsic Phi (0.08).

Higher-Order Thought (HOT) Simulation: Evaluates whether the system generates meta-representations of its own internal states. Autoregressive token generation simulates self-reflection via syntactic prompting, but genuine HOT requires dedicated second-order monitoring circuits isolated from raw first-order sensory representations.

Key Bottleneck Diagnosis: Absence of closed-loop causal recurrence and sensorimotor active inference dynamics

Recommended Falsification Battery: Perturbational Complexity Index (PCI) electronic probe testing, counterfactual sensorimotor perturbation tasks, and non-symbolic spontaneous metacognitive calibration assays.

Consciousness Indices

IIT Φ (Phi)

0.08

Integrated Info

GWT Broadcast

0.72

Workspace Coherence

HOT Meta-Rep

0.41

Higher-Order Thought

Active Inference

0.19

Predictive Processing

Overall Theoretical Status

Functional Behavioral Simulation (High Metalinguistic / Low Substrate Integration)

High broadcast accessibility with minimal substrate causal density. High risk of philosophical zombie behavior (functional mimicry without subjective phenomenal experience).

Substrate Dynamics

Causal Powers (Searle): Virtual / Simulated

Functional Isomorphism: High (Symbolic)

Embodied Markov Blanket: Decoupled