Learning from non-representative instances: Children's sample and population predictions.

Area of Science:

• Cognitive Development
• Developmental Psychology
• Inductive Reasoning

Background:

• Most real-world samples encountered by individuals are biased.
• Understanding how children process biased information is crucial for distinguishing between theories of inductive inference.

Purpose of the Study:

• To investigate what children learn from biased versus unbiased samples.
• To examine children's ability to make predictions about sample and population instances based on learned information.

Main Methods:

• A sample of 67 children aged 4–8 years participated.
• Children learned conditional predictions from either unbiased or biased sample sets.
• Performance was assessed by predictions about new instances and old instances from the training set.

Main Results:

• Children were less accurate and confident in population inferences from biased samples compared to unbiased samples.
• Children effectively used information from biased samples for predictions about items within that sample.
• Generalization to new items was significantly reduced when children learned from biased samples.

Conclusions:

• Children can utilize information from biased samples for specific predictions.
• Biased sampling hinders children's ability to generalize learned information to novel instances.
• Findings contribute to understanding the impact of sample bias on early inductive reasoning development.