Further perceptions of probability: In defence of associative models

Area of Science:

• Cognitive Psychology
• Behavioral Sciences
• Computational Neuroscience

Background:

• Research has focused on learning stable probabilities, with limited understanding of how humans track changing (nonstationary) probabilities.
• A debate exists between associative learning and hypothesis-testing models for nonstationary probability learning.

Purpose of the Study:

• To re-evaluate claims that hypothesis-testing models are necessary for explaining nonstationary probability learning.
• To propose and test a combined associative learning and evidence accumulation model.

Main Methods:

• Analysis of an experimental paradigm comprising probability tracking and change detection tasks.
• Development of a computational model integrating the delta learning rule and bounded evidence accumulation.
• Quantitative comparison against the hypothesis-testing model by Gallistel et al. (2014).

Main Results:

• The proposed combined model successfully accounts for qualitative patterns in the data.
• The combined model quantitatively outperforms the hypothesis-testing model.
• The experimental paradigm was shown to involve both estimation and decision-making components.

Conclusions:

• Nonstationary probability learning in humans can be explained by a combination of associative learning and bounded evidence accumulation.
• Existing data do not necessitate a novel hypothesis-testing model for this cognitive process.
• This finding supports a cumulative science approach by integrating established theories.