Bayesian confusions surrounding simplicity and likelihood in perceptual organization.

Area of Science:

• Cognitive Science
• Computational Neuroscience
• Psychology

Background:

• Perceptual organization models often invoke simplicity and likelihood principles.
• These principles, promoting efficiency and veridicality respectively, have been claimed to be equivalent.
• Existing models may show similar outcomes, particularly in common scenarios.

Purpose of the Study:

• To critically evaluate the claimed equivalence between the Occamian simplicity and Helmholtzian likelihood principles.
• To differentiate between subjective and objective probabilities in Bayesian frameworks for these principles.
• To clarify confusions arising from the dual role of regularity in perceptual organization.

Main Methods:

• Theoretical analysis of Bayesian models applied to perceptual organization.
• Distinction between subjective probabilities (Occamian) and objective probabilities (Helmholtzian).
• Contrast of complete and incomplete Occamian approaches to address perceptual confusions.

Main Results:

• The equivalence of Occamian and Helmholtzian principles is challenged due to a conflation of subjective and objective probabilities.
• Bayesian modeling requires distinct probability types for each principle.
• A dual role of regularity in perception contributes to conceptual confusion.

Conclusions:

• The Occamian simplicity and Helmholtzian likelihood principles are not fundamentally equivalent in perceptual organization.
• Accurate Bayesian modeling necessitates a clear distinction between subjective and objective probabilities.
• Resolving confusions requires a nuanced understanding of regularity's role and distinct theoretical frameworks.