Bidirectional influences of information sampling and concept learning.

Area of Science:

• Cognitive Science
• Computational Neuroscience
• Decision Science

Background:

• Current categorization models often neglect initial information encoding during decision-making.
• Human decision-makers selectively sample information in complex environments due to computational constraints.
• Selective attention plays a key role in modifying stimulus representations, accentuating relevant dimensions and compressing irrelevant ones.

Purpose of the Study:

• To develop an active sampling model that addresses limitations in current decision-making and categorization theories.
• To investigate how strategic information sampling influences knowledge representation and belief formation.
• To model the interplay between information utility assessment and environmental knowledge.

Main Methods:

• Development of the Sampling Emergent Attention (SEA) model, an active sampling framework.
• Modeling the interaction between two components: information source utility determination and environmental knowledge representation.
• Simulating sequential information sampling based on expected cost-benefit analysis.

Main Results:

• The SEA model demonstrates strategic information sampling, including search termination and adaptive path adjustment.
• The model replicates human-like decision-making behaviors, such as selective information processing.
• Bidirectional influences between sampling and learning can lead to belief divergence from reality, particularly when prioritizing exploitation over exploration.

Conclusions:

• The SEA model provides a computational framework for understanding active information sampling in decision-making.
• The findings highlight the importance of the interplay between knowledge and information seeking in shaping beliefs.
• The model's human-like failure modes underscore the potential for systematic biases in belief formation under certain conditions.