Can statistical learning bootstrap the integers?

Area of Science:

• Cognitive Science
• Developmental Psychology
• Computational Linguistics

Background:

• Children's acquisition of number words and cardinalities is a key developmental milestone.
• Piantadosi, Tenenbaum, and Goodman (2012) proposed a computational model for this learning process.
• The model utilizes statistical learning to establish the number-cardinality relation.

Purpose of the Study:

• To critically evaluate Piantadosi et al.'s (2012) model of children's number learning.
• To examine the model's claim of "Quinian bootstrapping."
• To assess the model's relevance and accuracy in simulating children's actual learning methods.

Main Methods:

• Analysis of the computational mechanisms within the Piantadosi et al. (2012) model.
• Comparison of the model's learning process with theories of conceptual change and bootstrapping.
• Evaluation of the model's starting assumptions (primitives) and learned representations.

Main Results:

• The model does not perform "Quinian bootstrapping" as it lacks conceptual discontinuity.
• Learning involves recombining primitives and statistical confirmation, aligning with Fodor's views on expressive power.
• The model's learned procedure and initial primitives differ from typical child development.

Conclusions:

• The model's learning mechanism is not analogous to true conceptual bootstrapping.
• The simulation's relevance to children's learning is questionable due to methodological differences.
• The model's partial understanding of integers allows for non-standard interpretations, highlighting potential ambiguities.