Drift in children's categories: when experienced distributions conflict with prior learning

Area of Science:

• Cognitive Psychology
• Developmental Psychology
• Machine Learning

Background:

• Psychological intuitions about category structure may not align with objective reality.
• Understanding how children resolve conflicts between innate biases and external information is crucial for developmental psychology.

Purpose of the Study:

• To investigate how young children respond to discrepancies between intuitive category structures and authoritative feedback.
• To examine the role of distributional information versus corrective feedback in children's learning and adaptation.

Main Methods:

• A semi-supervised learning paradigm was employed with 160 children aged 4–8 years.
• Children learned a one-dimensional decision criterion for classifying 'alien fruits' and then categorized new stimuli.
• The distribution of new stimuli was manipulated to create conflict with the initially learned boundary.

Main Results:

• Children adjusted their decision criteria to match the structure of the new stimuli, indicating unlearning of the original boundary.
• Younger children were more influenced by distributional information and less sensitive to corrective feedback.
• This suggests a developmental shift in how children weigh different sources of information during category learning.

Conclusions:

• Children demonstrate a capacity to adapt their category representations based on new evidence, even when it contradicts prior learning.
• Developmental differences exist in the ability to selectively attend to feedback versus distributional patterns.
• Findings highlight children's drive to form useful representations by integrating diverse informational cues.