A Spiking Neural Network Model of Elementary Self-Consciousness via Endogenous Default Mode Network Dynamics

Understanding self-referential cognition and baseline self-consciousness remains a fundamental challenge in computational neuroscience. This work proposes a large-scale computational model incorporating a 10,000-neuron spiking neural network (SNN) based on Izhikevich dynamics. The network is structured into two interacting subsystems: a sensory processing layer (5,000 regular-spiking cortical neurons) and an endogenous Default Mode Network (DMN) pacemaker subsystem (5,000 intrinsically bursting neurons). The DMN layer is modulated by continuous tonic currents reflecting ascending brainstem neuromodulation, maintaining intrinsic, autonomous bioelectric rhythms independent of external sensory input. To represent top-down cognitive modulation, synaptic weights are hierarchically structured such that DMN-to-network projections exceed sensory-level connections. Through numerical simulations using a modified two-step Euler integration scheme, the authors demonstrate "how endogenous pacemaker activity interacts with transient external sensory perturbations, providing an elementary mathematical framework for the emergence of a persistent, self-sustaining neural representation of 'Self'."

The model is formulated using Izhikevich's two-variable differential framework. The network comprises N = 10,000 spiking neurons divided into two functional subnetworks: the Sensory Processing Layer (i = 0 to 4,999), which models cortical excitatory pyramidal neurons exhibiting regular spiking (RS) dynamics with randomly drawn resetting parameters; and the Default Mode Network (DMN) Pacemaker Subsystem (i = 5,000 to 9,999), which models pacemaker-like bursting units where a core subset is assigned deterministic parameters to induce intrinsically bursting (IB) behavior. The network exhibits sparse connectivity with K = 1,000 random outgoing projections per neuron. Synaptic weights are defined by Sij = U (0,1) wij where the weight function follows a top-down hierarchical decision function: "wij = 0.6 j ≥ 5000 (DMN projections) / wj < 5000 (Sensory projections) =   0.4 j < 5,999 (Sensory projections) and w = 0.6."

The total input current to neuron i is defined as: I i (t) = I i, tonic + I i, ext + I i, syn +ηi (t). The Endogenous Tonic Current is applied continuously to the core DMN pacemaker neurons: Itonic = 7.0 pA if 5000 ≤ i ≤ 5999, otherwise 0.0. Exogenous Transient Stimuli are modeled as a stepcurrent applied to a subpopulation of sensory neurons during a defined temporal window: "I i, ext = 16.0 pA if 0 ≤ i < 1500, otherwise 0.0 for the time window 300 ms < t < 600 ms. Synaptic Current Integration models signal propagation as: I i, syn (t) = ∑ Sij (j v j (t) ≥ 30 mV)." Cognitive Noise is modeled as uncorrelated Gaussian N(0, σ 2) noise with σ = 3.

To quantify conscious experience and Qualia, the model adapts Integrated Information Theory (IIT). The scalar metric φ quantifies this degree of informational integration: φ measures the degree of dynamic functional coupling between the sensory layer and the Default Mode Network (DMN), capturing how external sensory perturbations modify and covary with the endogenous 'Self' dynamics. This is calculated by partitioning the active neurons into a Sensory submatrix and a DMN submatrix, constructing a partitioned block-diagonal covariance matrix ∑ Part, and quantifying φ as φ = ln (2 det(∑) / det(∑ Part)) when cross-network correlations are zero, indicating functional independence. Strong reciprocal covariance between sensory and DMN neurons drives the determinant ratio to be less than one, increasing φ and signaling the emergence of an integrated phenomenal state of consciousness or Qualia state.

Simulation results over a 900 ms trajectory show distinct transitions: Endogenous basal activity is observed during quiescent windows, e.g., t ∈ [150,300] ms, t ∈ [430, 480] ms, and t ∈ [650,780] ms, where core DMN pacemakers maintain persistent, asynchronous low-frequency firing driven by the 7.0 pA tonic current (Itonic). Four major synchronized population events occur around t ≈ 30 ms, t ≈ 330 ms, t ≈ 550 ms, and t ≈ 830 ms. During these transient periods, strong recurrent excitation within the DMN and top-down modulatory projections (wDMN = 0.6) recruit the sensory layer, driving global population firing rates to saturation (≈ 1000 Hz). Peak values of integrated information occur during sustained co-activation phases, with The highest peaks in integrated information (φ ≈ 12.5-15.0 bits) occur during the sustained co-activation phase of the bursts. Isolated states with φ dropping to zero are observed in quiescent periods, while pre-burst priming states (φ ≈ 1-2.5 bits) characterize the subtle functional coupling that begins prior to full global ignition. The discussion concludes that maximum values of φ (≈ 12.5-15.0 bits) emerge during transient population bursts when incoming sensory perturbations trigger widespread coactivation and phase synchrony across the sensory–DMN boundary, supporting the hypothesis that integrated phenomenal states (Qualia) emerge from the dynamic, non-linear coupling between internal selfreferential dynamics and sensory inputs. The paper acknowledges that this is a functional simulation of conscious dynamics and thus constitutes a computational artifact or a functional 'philosophical zombie' at the micro-scale, as it lacks subjective, first-person phenomenology despite its non-zero φ values. Future work includes porting the architecture onto neuromorphic hardware and incorporating Spike-Timing-Dependent Plasticity (STDP). The authors also plan to compare trajectories of φ (t) against empirical electrophysiological recordings from living human organoid networks for biological validation.

Acknowledgments: I would like to express my sincere gratitude to Google's Gemini AI for its invaluable collaboration throughout this research.

References include foundational works on IIT, DMN studies, and philosophical debates concerning consciousness (Chalmers, Searle). The paper cites Raichle et al. (2001), Raichle (2015), Steriade (2006), Varela et al. (2001), Tononi (2004, 2016), and Searle (1980, 1992). The model is implemented using Python 3.13 on a MacBook Air under macOS Tahoe 26.6.2 and uses a modified two-step Euler integration scheme with a fixed time step ∆t = 1.0 ms."

(Self-Correction/Final Review: Ensure the response is only the summary, detailed, and quoted.)

The model is formulated using Izhikevich's two-variable differential framework. The network comprises N = 10,000 spiking neurons divided into two functional subnetworks: the Sensory Processing Layer (i = 0 to 4,999), which models cortical excitatory pyramidal neurons exhibiting regular spiking (RS) dynamics with randomly drawn resetting parameters; and the Default Mode Network (DMN) Pacemaker Subsystem (i = 5,000 to 9,999), which models pacemaker-like bursting units where a core subset is assigned deterministic parameters to induce intrinsically bursting (IB) behavior. The network exhibits sparse connectivity with K = 1,000 random outgoing projections per neuron. Synaptic weights are defined by Sij = U (0,1) wij where the weight function follows a top-down hierarchical decision function: "wij = 0.6 j ≥ 5000 (DMN projections) / wj < 5,999 (Sensory projections) =  0.4 j < 5,999 (Sensory projections) and w = 0.6."

Simulation results over a 900 ms trajectory show distinct transitions: Endogenous basal activity is observed during quiescent windows, e.g., t ∈ [150,300] ms, t ∈ [430, 480] ms, and t ∈ [650,780] ms, where core DMN pacemakers maintain persistent, asynchronous low-frequency firing driven by the 7.0 pA tonic current (Itonic). Four major synchronized population events occur around t ≈ 30 ms, t ≈ 330 ms, t ≈ 550 ms, and t ≈ 830 ms. During these transient periods, strong recurrent excitation within the DMN and top-down modulatory projections (wDMN = 0.6) recruit the sensory layer, driving global population firing rates to saturation (≈ 1000 Hz). Peak values of integrated information occur during sustained co-activation phases, with The highest peaks in integrated information (φ ≈ 12.5-15.0 bits) occur during the sustained co-activation phase of the bursts. Isolated states with φ dropping to zero are observed in quiescent periods, while pre-burst priming states (φ ≈ 1-2.5 bits) characterize the subtle functional coupling that begins prior to full global ignition. The discussion concludes that maximum values of φ (≈ 12.5-15.0 bits) emerge during transient population bursts when incoming sensory perturbations trigger widespread coactivation and phase synchrony across the sensory–DMN boundary, supporting the hypothesis that integrated phenomenal states (Qualia) emerge from the dynamic, non-linear coupling between internal selfreferential dynamics and sensory inputs. The paper acknowledges that this is a functional simulation of conscious dynamics and thus constitutes a computational artifact or a functional 'philosophical zombie' at the micro-scale, as it lacks subjective, first-person phenomenology despite its non-zero φ values. Future work includes porting the architecture onto neuromorphic hardware and incorporating Spike-Timing-Dependent Plasticity (STDP)."

References include foundational works on IIT, DMN studies, and philosophical debates concerning consciousness (Chalmers, Searle). The model is implemented using Python 3.13 on a MacBook Air under macOS Tahoe 26.6.2 and uses a modified two-step Euler integration scheme with a fixed time step ∆t = 1.0 ms."

(End of Summary)

(Final Output Generation)

Understanding self-referential cognition and baseline self-consciousness remains a fundamental challenge in computational neuroscience. This work proposes a large-scale computational model incorporating a 10,000-neuron spiking neural network (SNN) based on Izhikevich dynamics. The network is structured into two interacting subsystems: a sensory processing layer (5,000 regular-spiking cortical neurons) and an endogenous Default Mode Network (DMN) pacemaker subsystem (5,000 intrinsically bursting neurons). The DMN layer is modulated by continuous tonic currents reflecting ascending brainstem neuromodulation, maintaining intrinsic, autonomous bioelectric rhythms independent of external sensory input. To represent top-down cognitive modulation, synaptic weights are hierarchically structured such that DMN-to-network projections exceed sensory-level connections. Through numerical simulations using a modified two-step Euler integration scheme, the authors demonstrate "how endogenous pacemaker activity interacts with transient external sensory perturbations, providing an elementary mathematical framework for the emergence of a persistent, self-sustaining neural representation of 'Self'."

Here are the specific improvements that can be made to AI systems, based on the findings and model presented in this paper:

1. Improve AI systems by incorporating a hierarchical, two-subsystem architecture mirroring biological cognition: A Sensory Processing Layer (for external input) coupled with an intrinsically rhythmic Default Mode Network (DMN) Pacemaker Subsystem for internal, self-referential processing.

2. Enhance the ability of AI to model and generate persistent, autonomous internal states (baseline self-consciousness) that are maintained independently of immediate external sensory input via continuous tonic neuromodulatory currents (modeled as fixed bioelectric rhythms).

3. Develop AI systems capable of modeling top-down cognitive modulation where internal, self-referential dynamics exert stronger control over lower-level sensory processing than the reverse (as reflected by the synaptic weight asymmetry: wDMN = 0.6 vs. wSensory = 0.4).

4. Improve qualitative understanding by integrating Integrated Information Theory (IIT) metrics into the AI's internal state representation, allowing it to quantify Qualia or integrated phenomenal states through a scalar metric (φ).

5. Enable AI to dynamically transition between distinct cognitive states—from quiescent/isolated baseline activity to pre-burst priming, and finally to peak integrated states—based on the degree of cross-network functional coupling and population firing rate synchronization.

6. Improve systems that require high levels of integrated awareness (e.g., complex decision-making or sustained attention) by designing them to operate in the peak integrated state where mutual information between internal self-referential dynamics and sensory inputs reaches its maximum capacity (φ ≈ 12.5–15.0 bits).

7. Increase AI robustness and learning capabilities by implementing Spike-Timing-Dependent Plasticity (STDP) mechanisms, allowing synaptic weights to evolve dynamically based on the history of co-activation between sensory and DMN units, enabling self-organization of cognitive structures over time.

8. For future implementation on specialized hardware, transition the underlying computational substrate from classical processing to neuromorphic chips (e.g., Intel Loihi), which can process continuous time dynamics and massive parallelism more closely aligned with biological spiking neural networks.

Understanding the neurobiological mechanisms underlying self-referential cognition and baseline self-consciousness remains a fundamental challenge in computational neuroscience. In this work, we propose a large-scale computational model incorporating a 10,000-neuron spiking neural network (SNN) based on Izhikevich dynamics. The network is structured into two interacting subsystems: a sensory processing layer (5,000 regular-spiking cortical neurons) and an endogenous Default Mode Network (DMN) pacemaker subsystem (5,000 intrinsically bursting neurons). The DMN layer is modulated by continuous tonic currents reflecting ascending brainstem neuromodulation, maintaining intrinsic, autonomous bioelectric rhythms independent of external sensory input. To represent top-down cognitive modulation, synaptic weights are hierarchically structured such that DMN-to-network projections exceed sensory-level connections. Through numerical simulations using a modified two-step Euler integration scheme, we demonstrate how endogenous pacemaker activity interacts with transient external sensory perturbations, providing an elementary mathematical framework for the emergence of a persistent, self-sustaining neural representation of "Self".