Children can solve Bayesian problems: the role of representation in mental computation

Area of Science:

• Cognitive Development
• Bayesian Reasoning
• Probabilistic Cognition

Background:

• Understanding children's capacity for complex reasoning is crucial.
• Bayesian reasoning, a sophisticated form of probabilistic inference, has been a focus of cognitive research.
• Previous research has not extensively explored eliciting Bayesian reasoning in children using specific data representations.

Purpose of the Study:

• To investigate whether children can perform Bayesian reasoning when information is presented using natural frequencies.
• To compare children's Bayesian reasoning performance across different age groups (fourth, fifth, and sixth graders).
• To examine the impact of data representation (probabilities vs. natural frequencies) on children's Bayesian reasoning.

Main Methods:

• Two studies were conducted with fourth, fifth, and sixth graders.
• Participants were presented with information in terms of either probabilities or natural frequencies.
• Performance was measured by the ability to estimate Bayesian posterior probability.

Main Results:

• Children's Bayesian reasoning ability was zero when information was presented as probabilities.
• When presented with natural frequencies, Bayesian reasoning increased from fourth to sixth grade (19% to 53%).
• Sixth graders' performance with natural frequencies matched adult performance with probabilities, but with significant individual differences.

Conclusions:

• The representation of information significantly impacts children's ability to perform Bayesian reasoning.
• Natural frequencies facilitate Bayesian reasoning in children more effectively than probabilities.
• Children's probabilistic reasoning, more than adults', is dependent on appropriate information representation, with non-Bayesian strategies persisting into adulthood.