Children's active-condition predictions closely align with an optimal Bayesian learner, supporting rational Bayesian inference as a model of children's belief revision¶

Claim · All claims
Evidence · 1 study · causal r3 · q2
Currency · newest study 2025 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? An optimal Bayesian model accurately predicted 80.4% of active-condition children's learning-phase responses, and was a strong predictor in a mixed-effects regression (β = 0.67, SE = 0.017, z = 38.91, p < 0.001). → Brod 2025

Evidence¶

Brod 2025¶

Brod, G., Holstein, E., Weindorf, L., Colantonio, J., Bonawitz, E. & Theobald, M. (2025). Do it yourself: discerning the effects of self-directed activity on conceptual learning. npj Science of Learning. https://doi.org/10.1038/s41539-025-00364-9

q2 · i? · causal · r3

Model comparison using active-condition responses: "the optimal Bayesian model accurately predicted 80.4% of the responses during the learning phase." A linear mixed-effects regression confirmed "the Bayesian model was a strong predictor of children's choices in the active prediction condition" (β = 0.67, SE = 0.017, z = 38.91, p < 0.001), with a non-significant random intercept.

"the optimal Bayesian model accurately predicted 80.4% of the responses during the learning phase"

Discussion¶

Related Claims¶

• Children integrate prior knowledge with new statistical evidence in a Bayesian way, overturning initially likely hypotheses — a broader claim this one bears on
• The Bayesian yoking procedure closely approximates individualized ideal learning paths, with no posttest or transfer difference between perfect and slightly mismatched children — related