Quantum formalism for cognitive psychology

1. Problem Statement

The paper addresses the challenge of modeling the dynamics of human cognitive states, specifically decision-making processes under uncertainty. While classical probability theory (Bayesian updating) is the standard for modeling rational decision-making, it faces limitations in explaining:

• Context effects: How the order of questions or information arrival alters outcomes (violations of the law of total probability).
• Entrenched misconceptions: Why individuals often fail to update their beliefs away from false narratives even when presented with new evidence (confirmation bias).
• Incompatibility: The inability of classical frameworks to handle situations where different decisions or "observables" cannot be simultaneously defined or measured.

The author posits that while quantum effects (like brain-level interference) are not necessarily required, the mathematical formalism of quantum theory (Hilbert spaces, projection postulates, and tensor products) offers a more robust and geometric framework for characterizing the dynamics of the "state of mind."

2. Methodology

The author employs a mathematical framework mapping cognitive processes onto quantum mechanical structures:

• State of Mind Representation: The cognitive state regarding $N$ alternatives is represented as a vector $|\psi\rangle$ in a high-dimensional Hilbert space $\mathcal{H}$. The probability $p_k$ of choosing alternative $k$ is the square of the amplitude: $p_k = |\psi_k|^2$.
• Tensor Product Structure: For multiple decisions, the state space is the tensor product of individual decision spaces ($\mathcal{H} = \otimes \mathcal{H}_l$). This allows for the modeling of entanglement (statistical dependencies between decisions) and product states (independent decisions).
• Observables: Choices are modeled as Hermitian operators (observables) $\hat{X}$. The eigenvalues label the alternatives.
• Information Acquisition (Dynamics): The change in the state of mind upon receiving information is modeled using the von Neumann-Lüders projection postulate.
  • Information is modeled as a noisy signal: $\hat{\xi} = \hat{X} + \hat{\epsilon}$, where $\hat{\epsilon}$ represents environmental noise.
  • Upon observing a value $\xi$, the state collapses to a Lüders state, which is the orthogonal projection of the initial state onto the subspace consistent with the observation.
• Continuous Time Dynamics: The paper extends this to sequential information arrival using a time-series model where noise is a Brownian motion ($B_t$). This leads to a stochastic differential equation (SDE) for the state vector, driven by an "innovations process" ($dW_t$).
• Geometric Projection: The dynamics are analyzed by projecting the Hilbert space vector onto a Projective Hilbert Space (Real Projective Space), eliminating the irrelevant global phase/scaling factor.

3. Key Contributions

A. Equivalence of Projection and Bayesian Updating

The paper demonstrates that the von Neumann-Lüders projection postulate, when applied to a noisy information channel, yields the Bayes formula for conditional probabilities.

• The Lüders state $|\psi(\xi)\rangle$ corresponds to the state where the coefficients are the posterior probabilities derived from Bayes' theorem.
• This provides a geometric interpretation of Bayesian updating: it is the projection of the initial state onto the subspace defined by the new evidence, minimizing the distance (Bhattacharyya distance) to the initial state.

B. The Free Energy Principle and Uncertainty Minimization

The author proves that the Lüders projection is the unique dynamical evolution that minimizes the expected future uncertainty (surprise).

• By minimizing the conditional variance of the choice observable, the system adheres to the Free Energy Principle (widely used in neuroscience), which posits that biological systems act to minimize surprise.
• This establishes that the "tendency" of the mind toward low-entropy states is a natural consequence of information-theoretic efficiency, not just a biological imperative.

C. Geometric Explanation of "Tenacious Bayesian" Behavior

By analyzing the dynamics on the Projective Hilbert Space, the paper reveals that the drift term of the stochastic equation is the negative gradient of the variance (uncertainty).

• Result: States with zero uncertainty (eigenstates) act as attractors.
• Implication: If a person's mind is close to a "false" certainty (a false eigenstate), the dynamics naturally resist moving away from it because doing so would require an increase in uncertainty (entropy) before a decrease can occur. This mathematically explains confirmation bias and the difficulty of correcting false beliefs using rational Bayesian updating alone.

D. Resolution of Order-Dependence via Incompatible Observables

The paper addresses the limitation of classical reasoning where the order of questions matters (e.g., the Clinton/Gore honesty experiment).

• Classical Failure: If all observables commute, the Law of Total Probability holds, and order does not matter.
• Quantum Solution: By introducing incompatible (non-commuting) observables, the model allows for context-dependent probabilities.
• Mechanism: Measuring observable $\hat{Y}$ first changes the state such that the subsequent measurement of $\hat{X}$ yields different statistics than if $\hat{X}$ were measured first. This successfully models empirical violations of the Law of Total Probability.

E. Escape from False Attractors

A critical contribution is showing how incompatible observables can rescue a cognitive system trapped in a false attractor.

• In a purely classical (commuting) framework, a person stuck in a false belief cannot escape via rational updating without a massive, unlikely noise event.
• In the quantum framework, introducing information about an incompatible observable (a different "question" or perspective) can rotate the state vector out of the false attractor's basin of attraction, allowing the system to eventually converge to the truth.

4. Results

• Derivation of Stochastic Dynamics: The paper derives the stochastic differential equation for the state of mind:
  $$d|\psi(\xi_t)\rangle = -\frac{1}{8}(\hat{X} - \langle\hat{X}\rangle_t)^2 |\psi(\xi_t)\rangle dt + \frac{1}{2}(\hat{X} - \langle\hat{X}\rangle_t) |\psi(\xi_t)\rangle dW_t$$
  This equation explicitly links the drift (uncertainty reduction) and diffusion (information arrival) to the variance of the observable.
• Lindblad Equation: The averaged reduced density matrix obeys a Lindblad equation, confirming consistency with open quantum system dynamics.
• Quantitative Fit: The model successfully fits empirical data regarding question-order effects (e.g., the Clinton/Gore study) by tuning the eigenvalue differences and the angle of incompatibility ($\phi$) between observables.

5. Significance

• Theoretical Unification: The paper bridges cognitive psychology, information theory, and quantum formalism, showing that quantum-like structures are intrinsic to any probabilistic system modeled on a Hilbert space, regardless of whether the brain is physically quantum.
• New Mechanism for Bias Correction: It offers a novel explanation for how humans can overcome deep-seated misconceptions (confirmation bias). It suggests that changing the context or perspective (introducing incompatible observables) is mathematically necessary to escape false beliefs, rather than just presenting more of the same evidence.
• Alternative to Classical Reasoning: It challenges the sufficiency of classical probability for modeling human cognition, particularly in scenarios involving context, order effects, and the dynamics of belief formation.
• Free Energy Principle: It provides a rigorous mathematical derivation showing that the Free Energy Principle (minimizing surprise) emerges naturally from the projection postulate of quantum mechanics.

In summary, Brody argues that the quantum formalism is not merely an analogy but a necessary and effective tool for modeling the dynamics of cognitive states, particularly for explaining phenomena that classical probability theory fails to capture, such as the persistence of false beliefs and the order-dependence of human judgment.