Quantum-Like Bayesian Networks for Modeling Decision Making.

Area of Science:

• Cognitive Science
• Quantum Physics
• Artificial Intelligence

Background:

• Human decision-making often violates classical probability laws, as seen in the Sure Thing Principle paradoxes.
• Existing models struggle to generalize beyond simple scenarios or explain these paradoxical findings.
• Quantum-like models offer potential but face challenges with parameter complexity.

Purpose of the Study:

• To propose a novel Quantum-Like Bayesian Network (QLBN) for probabilistic inference.
• To address the limitations of classical probability in explaining paradoxical human choices.
• To develop a generalizable and predictive model for decision-making.

Main Methods:

• Replaced classical probabilities with quantum probability amplitudes within a Bayesian Network framework.
• Introduced a similarity heuristic for automatic quantum parameter fitting using vector similarities.
• Developed a non-parametric method for statistical inference estimation.

Main Results:

• The QLBN successfully accommodates violations of classical probability theory.
• The model demonstrated generalizability and predictive accuracy on empirical data.
• Tested scenarios included the Prisoner's Dilemma and Two Stage Gambling games.

Conclusions:

• The proposed QLBN offers a simpler, more general, and predictive alternative to existing quantum-like and dynamic models.
• This approach provides a robust framework for understanding and predicting human decision-making, even when it deviates from classical norms.