quantum-confirmation-bias-adaptive

name	quantum-confirmation-bias-adaptive
description	Quantum probability framework for understanding confirmation bias as optimal evidence selection - square-root probability spaces, matrix-valued observations, and evolutionary advantages in sequential hypothesis testing.
category	quantum
tags	["quantum-probability","confirmation-bias","hypothesis-testing","active-inference","decision-theory"]
trigger_words	["quantum confirmation bias","adaptive confirmation","square-root probability","quantum hypothesis testing","active quantum inference","optimal evidence selection","confirmation bias rationality"]
source	arXiv:2606.23325

The Adaptive Nature of Confirmation Bias (Quantum Probability Framework)

Overview

Confirmation bias is formulated on the space of square-root probabilities using quantum probability structures. Observations are modeled by matrices rather than random variables on a probability space. In binary hypothesis testing, the optimal evidence choice that minimizes expected error probability leads to confirmation bias - revealing this as a rational strategy with evolutionary advantages.

Core Methodology

Square-Root Probability Space

Instead of classical probability space (Omega, F, P), work on the space of square-root probabilities:

• States are vectors psi in C^n with |psi_i|^2 = p_i
• Observations are matrices (operators) rather than random variables
• This enables quantum-like interference effects in belief updating

Optimal Evidence Selection

In sequential binary hypothesis testing (H0 vs H1):

1. At each step, the decision maker chooses which evidence to sample
2. The optimal choice minimizes expected error probability
3. This optimal choice inherently produces confirmation bias

Evolutionary Advantages

Two remarkable advantages emerge:

1. Minimal Memory: Decision maker requires only the smallest memory capacity
2. Exponential Error Reduction: Error probability decreases exponentially in sample size

Active Inference Connection

The framework connects to active inference where the decision maker seeks evidence providing maximum information. The resulting optimal evidence agrees with the one obtained by minimizing error probability.

Implementation Patterns

Pattern 1: Square-Root Probability Representation

import numpy as np

class QuantumProbabilityState:
    """State represented as square-root probability amplitude vector."""
    
    def __init__(self, amplitudes):
        self.psi = np.array(amplitudes, dtype=complex)
        self.psi /= np.linalg.norm(self.psi)  # Normalize
    
    def probabilities(self):
        return np.abs(self.psi)**2
    
    def observe(self, measurement_matrix):
        """Apply observation (matrix-valued measurement)."""
        # Born rule: P(outcome) = |<outcome|M|psi>|^2
        new_state = measurement_matrix @ self.psi
        new_state /= np.linalg.norm(new_state)
        return QuantumProbabilityState(new_state)

Pattern 2: Optimal Evidence Selection

def optimal_evidence_selection(prior, evidence_options, hypothesis_models):
    """Select evidence that minimizes expected error probability."""
    best_evidence = None
    min_error = float('inf')
    
    for evidence in evidence_options:
        # Expected error after observing this evidence
        expected_error = 0
        for h_idx, model in enumerate(hypothesis_models):
            likelihood = model.likelihood(evidence)
            posterior = bayes_update(prior, evidence, model)
            error = min(posterior)  # Error = probability of wrong hypothesis
            expected_error += prior[h_idx] * error
        
        if expected_error < min_error:
            min_error = expected_error
            best_evidence = evidence
    
    return best_evidence

Pattern 3: Confirmation Bias Emergence

def confirmation_bias_simulation(prior, evidence_stream, num_steps):
    """Simulate how optimal evidence selection produces confirmation bias."""
    state = QuantumProbabilityState([np.sqrt(prior[0]), np.sqrt(prior[1])])
    bias_history = []
    
    for step in range(num_steps):
        # Choose evidence that confirms current belief
        current_belief = state.probabilities()
        favored_hypothesis = np.argmax(current_belief)
        
        # Optimal evidence tends to confirm current belief
        evidence = select_confirmatory_evidence(favored_hypothesis)
        state = state.observe(evidence)
        
        bias_history.append(state.probabilities())
    
    return bias_history

Key Results

1. Confirmation Bias is Rational: Optimal evidence selection inherently produces confirmation bias
2. Minimal Memory: Only O(1) memory capacity needed for optimal sequential testing
3. Exponential Convergence: Error probability decreases as O(exp(-n)) in sample size
4. Active Inference Agreement: Maximum information seeking agrees with error minimization

Practical Implications

1. Confirmation bias should not be viewed as purely irrational
2. In sequential decision making, confirmatory evidence can be optimal
3. The tradeoff: faster convergence but risk of persistent wrong beliefs
4. Active inference provides complementary justification

Activation

Use this skill when:

• Analyzing confirmation bias from information-theoretic perspective
• Designing sequential hypothesis testing protocols
• Building agents with optimal evidence selection
• Studying the rationality of cognitive biases
• Working with quantum probability models of cognition
• Understanding active inference and information seeking