---
title: "Epistemic Calibration Across the Knowingness Quad: A Dual-Process Dynamical Model of the Dunning-Kruger Crossover and Affective Trait Gain"
subtitle: "Why Actual Ability Only Equals Perceived Ability at Proficiency—and How Subconscious Automation, Exposure Volume, and Big-Five Gain Parameters Drive the Perfectionist's Trap and Expert Underscoring"
author:
  - name: "Trent R. Carter"
    affiliation: "TrueSynthesis Inc. / Independent Research, Indialantic, FL"
    url: "https://trentcarter.com"
date: "2026-11-04"
format:
  pdf:
    documentclass: article
    classoption: [twocolumn]
    geometry: [margin=0.75in]
    number-sections: true
    colorlinks: true
  html:
    theme: cosmo
    toc: true
    toc-depth: 3
    code-fold: true
    self-contained: true
---

## Abstract

Metacognitive miscalibration—commonly referred to as the Dunning-Kruger effect—is frequently caricatured in popular literature as a single static curve featuring an early peak of inflated confidence followed by a sharp valley of despair. However, empirical percentile data and operational skill-acquisition trajectories reveal a dual-curve topology in which Perceived Ability ($P$) and Actual Ability ($A$) evolve monotonically at different rates, intersecting only at a specific mid-high competence threshold. In this paper, we unify the Dunning-Kruger dual-curve framework with the four-stage **Knowingness Quad** taxonomy (*Novice, Competent, Proficient, Expert*) by modeling skill acquisition as a two-bit **Gray-Code Epistemic State Machine** across Conscious ($C_k \in \{0,1\}$) and Sub-Conscious ($S_k \in \{0,1\}$) processing: $[0,0] \to [1,0] \to [1,1] \to [0,1]$. We prove structurally why the calibration crossover ($P = A$) occurs exclusively within the **Proficient** ($[1,1]$) regime at $c^* \approx 68.0$, and demonstrate that the transition to **Expert** ($[0,1]$)—where execution offloads from conscious effort to subconscious automation ($C_k: 1 \to 0$)—induces a negative effort-heuristic bias that causes top-quartile performers to systematically underscore their relative percentile ($P < A$).

Furthermore, we extend this population baseline into a coupled stochastic dynamical system modeled as an **Asymmetric Diode Rectifier and Leaky RC Integrator**. By treating Conscientiousness and Talent ($C \cdot T$) as horizontal state velocity and Extraversion ($\alpha_E$) and Neuroticism ($\alpha_N$) as asymmetric reward and threat Kalman filter gains on perceived competence, we partition the competence phase space into four behavioral phenotypes (*Arrogant, Insecure, Humble, Confident*). Crucially, this formulation provides a formal mathematical derivation for two pervasive behavioral phenomena: (1) **Exposure Volume ($V$) as a Bayesian Damping Force**, proving why real-world trial volume is the only mechanism capable of collapsing both low-skill arrogance and high-skill insecurity onto the $P = A$ reality diagonal; and (2) **The Perfectionist's Trap ($C \times \alpha_N$ Coupling)**, explaining why elite practitioners who leverage high neurotic error-detection ($\alpha_N$) and high conscientiousness ($C$) to accelerate their objective mastery into the 95th+ percentile simultaneously trap their internal self-perception in chronic self-doubt unless task-error auditing is decoupled from identity evaluation.

---

## 1. Introduction: Beyond the "Mount Stupid" Caricature

When Perceived Ability $P(c)$ and Actual Ability $A(c)$ are plotted simultaneously against objective domain competence $c \in [0, 100]$, both curves increase monotonically. As established in the March 2020 *Dunning-Kruger Effect + Knowingness Quad* formulation (Carter, 2020), the core phenomenon is a differential slope and phase-lag problem between two coupled trajectories:

1. **Actual Ability ($A(c)$):** Starts near the floor ($A \approx 10$ in the Novice regime) and exhibits convex accelerating growth across competence tiers before approaching asymptotic mastery ($A \to 98$) in the Expert regime.
2. **Perceived Ability ($P(c)$):** Starts with a high baseline offset ($P \approx 55$) due to the absence of evaluative meta-knowledge, rises with a shallow slope ($\frac{dP}{dc} < \frac{dA}{dc}$) across the middle regimes, intersects the Actual Ability curve at a single interior fixed point ($c^* \approx 68.0$) before falling beneath it ($P < A$) at the top end of the distribution.

> *"It is not until one is Proficient that their Actual ability is equal to their Perceived Ability."* (Carter, 2020)

![The Unified Dunning-Kruger + Knowingness Quad Master Diagram showing the 2020 baseline curves, the Proficient Crossover at $c^* \approx 68.0$, and the Big-Five trait envelopes.](figures/fig1_master_quad.png){#fig-master width=100%}

---

## 2. Cognitive Architecture: The 2-Bit Gray-Code Epistemic State Machine

Human skill execution and self-evaluation rely on two neuroanatomically distinct processing streams:

* **The Conscious Bus ($C_k \in \{0, 1\}$):** Explicit, declarative, working-memory-intensive rule processing and active error monitoring (dlPFC and ACC).
* **The Sub-Conscious Bus ($S_k \in \{0, 1\}$):** Implicit, procedural schema automation, motor chunking, and fluid pattern recognition (basal ganglia, SMA, and cerebellum).

Mapping **Don't Know** to $0$ and **Know** to $1$, human competence progression across the four regimes (*Novice, Competent, Proficient, Expert*) follows a strict **2-bit reflected binary Gray Code**:

$$\mathbf{k}(c) = \begin{bmatrix} C_k(c) \\ S_k(c) \end{bmatrix} : \quad \begin{bmatrix} 0 \\ 0 \end{bmatrix}_{\text{Novice}} \longrightarrow \begin{bmatrix} 1 \\ 0 \end{bmatrix}_{\text{Competent}} \longrightarrow \begin{bmatrix} 1 \\ 1 \end{bmatrix}_{\text{Proficient}} \longrightarrow \begin{bmatrix} 0 \\ 1 \end{bmatrix}_{\text{Expert}}$$

| Competence Regime ($c$) | Conscious ($C_k$) | Sub-Conscious ($S_k$) | Gray Code $[C_k, S_k]$ | Epistemic State Label | Baseline Actual ($A$) | Baseline Perceived ($P$) | Signed Error ($\Delta = P - A$) |
| :--- | :--- | :--- | :--- | :--- | :--- | :--- | :--- |
| **I. Novice** ($0\text{–}25$) | Don't Know | Don't Know | $[0, 0]$ | *Limited Knowledge* | $10 \to 25$ | $55 \to 58$ | $+45 \to +33$ |
| **II. Competent** ($25\text{–}50$) | Know | Don't Know | $[1, 0]$ | *Some Knowledge* | $25 \to 52$ | $58 \to 63$ | $+33 \to +11$ |
| **III. Proficient** ($50\text{–}75$) | Know | Know | $[1, 1]$ | *Solid Knowledge* | $52 \to 79$ | $63 \to 72$ | $+11 \to \mathbf{0} \to -7$ |
| **IV. Expert** ($75\text{–}100$) | Don't Know | Know | $[0, 1]$ | *Understanding* | $79 \to 98$ | $72 \to 83$ | $-7 \to -15$ |

### 2.1 The Information-Theoretic Cognitive Bottleneck: Unit Hamming Distance Constraint ($d_H = 1$)

Adjacent states in a reflected Gray code differ by exactly one bit ($d_H(s_i, s_{i+1}) = 1$). In human neurobiology, this reflects the working-memory capacity bottleneck (Baddeley, 1992; Miller, 1956). Basal ganglia schemas cannot compile *de novo* without conscious staging ($S_k: 0 \to 1$ requires prior rehearsal while $C_k = 1$). Therefore, the diagonal shortcut $[0,0] \to [1,1]$ ($d_H = 2$) is biologically forbidden. Conversely, once schemas compile into fluid intuition ($S_k = 1$), maintaining prefrontal rule-tracking is metabolically inefficient. Thus, expertise offloads the prefrontal monitor ($C_k: 1 \to 0$), transitioning to the terminal $[0,1]$ state.

![The 2-Bit Reflected Gray-Code Epistemic State Machine and Cognitive Information Bottleneck.](figures/fig2_gray_code_fsm.png){#fig-fsm width=100%}

---

## 3. Mathematical Derivation of the Proficient Crossover and Expert Underscoring

Define the **Epistemic Calibration Error** as the signed residual $\Delta_0(c) \equiv P_0(c) - A(c)$.

**Theorem 1 (Existence and Uniqueness of Proficient Calibration Point):** *There exists a unique competence level $c^* \in (50, 75)$ such that Actual Ability equals Perceived Ability: $P_0(c^*) = A(c^*)$. Furthermore, $c^* \approx 68.0$.*

*Proof.* Let $c$ be parametrized over $[0, 100]$. Both $A(c)$ and $P_0(c)$ are continuous and strictly increasing monotonic functions. At $c = 50$: $\Delta_0(50) = 63.0 - 52.0 = +11.0 > 0$. At $c = 75$: $\Delta_0(75) = 72.0 - 79.0 = -7.0 < 0$. By Bolzano's Intermediate Value Theorem, there exists at least one $c^* \in (50, 75)$ where $\Delta_0(c^*) = 0$. Because Actual Ability exhibits convex acceleration ($\frac{d^2A}{dc^2} > 0$) while Perceived Ability exhibits concave dampening ($\frac{d^2P}{dc^2} < 0$), $\frac{d\Delta_0}{dc} < 0$ strictly on $[50, 75]$. By strict monotonicity, the root $c^*$ is unique. Interpolation of the empirical baseline yields $c^* \approx 68.03$ and $A(c^*) = P_0(c^*) \approx 68.48$. $\blacksquare$

### 3.1 The Effort-Heuristic Inversion and Expert Underscoring

At the transition from Proficient ($[1, 1]$) to Expert ($[0, 1]$), the conscious monitoring bit flips OFF ($\Delta C_k = -1$). In subjective metacognition, self-assessed difficulty and task value are strongly coupled to prefrontal metabolic effort (the *Effort Heuristic*; Kruger et al., 2004; Kahneman, 2011). Because automated schemas feel subjectively effortless to the expert, their internal feedback loop misinterprets this absence of cognitive friction as evidence that the task is trivial ('anyone could do this'). Consequently, self-percentile evaluation plateaus or drops ($P(95) \approx 83$), creating an underscoring deficit of $\Delta_0 \approx -15$ points.

![Epistemic Calibration Error Residual and Gray-Code State Switching.](figures/fig3_gray_code_residual.png){#fig-residual width=100%}

---

## 4. Phase-Space Topology & The Four Behavioral Phenotypes

Projecting Actual Ability ($A$) and Perceived Ability ($P$) onto an orthogonal 2D phase plane partitions behavior into four operational zones separated by $A_{\text{crit}} = 50$ and the calibration boundary $P_{\text{cal}}(A)$:

1. **Arrogant ($A < 50, P > P_{\text{cal}}$):** Low skill paired with inflated self-perception; $C_k = 0$ blinds the agent to error.
2. **Insecure ($A < 50, P < P_{\text{cal}}$):** Prefrontal awakening ($C_k = 1$) before schema compilation ($S_k = 0$) induces acute self-doubt.
3. **Humble / Underscoring ($A \ge 50, P < P_{\text{cal}}$):** High objective competence paired with understated self-perception via the negative effort heuristic.
4. **Confident / Calibrated ($A \ge 50, P \ge P_{\text{cal}}$):** Reality-backed self-perception achieved through high exposure volume.

---

## 5. The Cognitive Hardware Model: Asymmetric Diode Rectifier & Leaky RC Integrator

In Cybernetic Big-Five Theory (DeYoung, 2015), Extraversion ($E$) and Neuroticism ($N$) represent internal forward gains on the brain's reward and threat prediction-error channels:

* **Extraversion ($\alpha_E$):** Forward gain across an ideal diode on the positive half-wave: $v^+(t) = \alpha_E \cdot \max(0, e(t))$.
* **Neuroticism ($\alpha_N \cdot C_k$):** Forward gain across an inverted diode on the negative half-wave, gated by a FET switch representing the Conscious bit: $v^-(t) = \alpha_N \cdot C_k \cdot \min(0, e(t))$.
* **Self-Perception ($P(t)$):** The smoothed DC bias voltage across a leaky RC capacitor tank damped by Exposure Volume ($V$):

$$\tau_V \frac{dP(t)}{dt} + P(t) = A + (1 - V) \cdot \left[ \alpha_E \cdot \max(0, e(t)) + \alpha_N \cdot C_k \cdot \min(0, e(t)) \right]$$

![Cognitive Hardware Signal-Processing Circuit Model and DSO Waveforms.](figures/fig4_rectifier_circuit_schematic.png){#fig-circuit width=100%}

---

## 6. Asymmetric Bayesian Kalman Filter Formulation & Reality Variance Collapse

Let true skill evolve via discrete state equation $x_{k+1} = x_k + w_k$, while observed subjective performance is perturbed by measurement noise $v_k \sim \mathcal{N}(0, R_k)$, where $R_k = R_0 \cdot \frac{1 - V}{V}$. The asymmetric innovation gain is:

$$K_k(\tilde{y}_k) = K_0 \cdot \left[ 1 + \alpha_E \cdot \mathbb{I}(\tilde{y}_k \ge 0) + \alpha_N \cdot C_k \cdot \mathbb{I}(\tilde{y}_k < 0) \right]$$

**Theorem 2 (Bayesian Variance Collapse):** *As Exposure Volume approaches unity ($V \to 1.0$), the observation noise covariance collapses to zero ($R_k \to 0$). Consequently, the posterior self-perception distribution converges in mean and variance onto objective ability $A$, regardless of affective trait skew ($\alpha_E, \alpha_N$).*

![Asymmetric Bayesian Kalman Filter Distribution Evolution across Exposure Volumes.](figures/fig5_bayesian_kalman_distributions.png){#fig-kalman width=100%}

---

## 7. Empirical Domain Case Studies: Four Phenotypes in Action

To demonstrate the real-world operational predictive power of the model, we trace four discrete domain case studies across 40 trials:

1. **Case A: Software Systems Architect (The Perfectionist's Trap):** High $C$ (0.92), High $N$ (0.88), Low $E$ (0.22). $A \to 94\%$, but $P \to 68\%$ ($\Delta = -26$). High $\alpha_N$ fuels code quality but depresses internal self-worth.
2. **Case B: Crypto / Derivative Trader (Unskilled Arrogance):** Low $C$ (0.25), Low $N$ (0.15), High $E$ (0.92). $A = 22\%$, but $P = 86\%$ ($\Delta = +64$). High reward sensitivity with unawakened conscious evaluation.
3. **Case C: Concert Pianist & Improviser (Subconscious Hand-off):** High $C$ (0.85), Mod $N$ (0.50), Mod $E$ (0.60). At trial 22, transitions $[1,1] \to [0,1]$. $A = 96\%$, $P = 81\%$ ($\Delta = -15$). Effortless execution discounted as ordinary.
4. **Case D: Laparoscopic Surgical Fellow (High-Volume Bayesian Calibration):** High $C$ (0.95), Balanced traits, High Volume ($V \to 0.95$). $A = 92.4\%, P = 91.1\%$ ($\Delta \approx 0$). Immense simulation volume collapses affective bias onto reality.

![Empirical Domain Case Studies Across Four Phenotypic Quadrants.](figures/fig6_domain_case_studies.png){#fig-domains width=100%}

---

## 8. Coupled Dynamical Trajectory Equations Across Discrete Trials

In a discrete-trial learning environment ($t = 1, \dots, N$):

$$A_{t+1} = A_t + \gamma \cdot (C \cdot T) \cdot \left( \frac{100 - A_t}{100} \right) \cdot \left[ 1 + \beta_{\text{debug}} \cdot \alpha_N \cdot \mathbb{I}(R_t^- = 1) \right]$$

$$P_{t+1} = P_t + \frac{1}{\tau_V} \cdot \left[ \left( A_t + (1 - V) \cdot v_{\text{rect}}(t) \right) - P_t \right]$$

![40-Trial Coupled Dynamical Trajectory Simulation.](figures/fig4_dynamical_trials.png){#fig-trials width=100%}

---

## 9. Diagnostic Decision Engine: The Epistemic Calibration Index (ECI)

We define the **Epistemic Calibration Index (ECI)**:

$$\text{ECI} \equiv 1 - \frac{|P - A|}{\max(P, A)} \in [0, 1]$$

An individual is calibrated when $\text{ECI} \ge 0.90$. When $\text{ECI} < 0.75$, leaders apply the diagnostic flowchart below:

![The Epistemic Calibration Diagnostic Decision Engine.](figures/fig8_diagnostic_flowchart.png){#fig-flowchart width=100%}

---

## 10. Five Core Clinical, Personal, and Organizational Frameworks

1. **Exposure Volume ($V$) as the Universal Antidote to Miscalibration:** Increasing trial volume $V$ multiplies the affective error wedge by $(1 - V)$, forcing $P \to A$. Validates ERP as a Bayesian variance-reduction operator.
2. **The 'Perfectionist's Trap' Coupling ($C \times \alpha_N$)—Decoupling Task Fuel from Identity:** Use $\alpha_N$ strictly as a task debugging sensor; gate it off before identity integration.
3. **The $[1,0]$ Transition Diagnostic—Reframing Acute Self-Doubt as Competence Growth:** Acute insecurity means the Conscious bit flipped ON ($C_k: 0 \to 1$). Motor schema compilation ($S_k$) naturally lags.
4. **The Effort-Heuristic Audit—De-biasing Expert Underscoring ($[0,1]$):** In the Expert regime, Conscious tracking flips OFF ($C_k: 1 \to 0$). Ease is proof of subconscious compilation, not mediocrity.
5. **Asymmetric Gain Calibration—Applying a Personal Gauge Offset:** Low-E / High-N practitioners must add a $+20$ percentile offset once $C_k = 1$; High-E / Low-N practitioners must subtract $-20$ and mandate external red-team verification.

---

## 11. Conclusion & Philosophical Synthesis

Metacognitive calibration is a dynamic equilibrium between prefrontal conscious auditing, basal ganglia procedural compilation, affective prediction-error gains, and external exposure volume. By reframing the Dunning-Kruger effect as a coupled Gray-code dynamical system, we provide a unified mathematical and operational language that honors the empirical data, explains why crossover occurs exclusively at Proficiency, and provides actionable tools to liberate high performers from chronic imposter doubt while protecting organizations from uncalibrated arrogance.

---

## References

1. Baddeley, A. (1992). Working memory. *Science*, 255(5044), 556–559.
2. Carter, T. R. (2020). *The Knowingness Quad and Dunning-Kruger Dual-Curve Formulation*. Indialantic, FL: TrueSynthesis Research Archives.
3. Carter, T. R. (2026). Epistemic Calibration Across the Knowingness Quad: A Dual-Process Dynamical Model of the Dunning-Kruger Crossover and Affective Trait Gain. *TrueSynthesis Technical Papers*. https://trentcarter.com/papers/knowingness-quad
4. DeYoung, C. G. (2015). Cybernetic Big Five Theory. *Journal of Research in Personality*, 56, 33–52.
5. Duckworth, A. L., Peterson, C., Matthews, M. D., & Kelly, D. R. (2007). Grit: Perseverance and passion for long-term goals. *Journal of Personality and Social Psychology*, 92(6), 1087–1101.
6. Ericsson, K. A., Krampe, R. T., & Tesch-Römer, C. (1993). The role of deliberate practice in the acquisition of expert performance. *Psychological Review*, 100(3), 363–406.
7. Fitts, P. M., & Posner, M. I. (1967). *Human Performance*. Belmont, CA: Brooks/Cole.
8. Friston, K. (2010). The free-energy principle: A unified brain theory? *Nature Reviews Neuroscience*, 11(2), 127–138.
9. Gray, F. (1953). Pulse code communication. *U.S. Patent 2,632,058*.
10. Kahneman, D. (2011). *Thinking, Fast and Slow*. New York: Farrar, Straus and Giroux.
11. Kruger, J., & Dunning, D. (1999). Unskilled and unaware of it: How difficulties in recognizing one's own incompetence lead to inflated self-assessments. *Journal of Personality and Social Psychology*, 77(6), 1121–1134.
12. Kruger, J., Wirtz, D., Boven, L. V., & Altermatt, T. W. (2004). The effort heuristic. *Journal of Experimental Social Psychology*, 40(1), 91–98.
13. Miller, G. A. (1956). The magical number seven, plus or minus two: Some limits on our capacity for processing information. *Psychological Review*, 63(2), 81–97.
14. Schneider, W., & Shiffrin, R. M. (1977). Controlled and automatic human information processing: I. Detection, search, and attention. *Psychological Review*, 84(1), 1–66.
15. Sweller, J. (1988). Cognitive load during problem solving: Effects on learning. *Cognitive Science*, 12(2), 257–285.