Downloads & Free Reading Options - Results

Product reviews of spatial toys—blocks, puzzles, shape sorters, and building sets—offer a unique window into how people value spatial play in early childhood. These reviews contain millions of unstructured text data points that reflect people's underlying beliefs and priorities. When someone writes that a puzzle "helped with problem-solving skills," states building blocks were "worth every penny," mentions their child "loves it when they label the shape words," or says they purchased a shape sorter to help their child's motor skills improve, they're sharing their beliefs about why it's important to provide children with access to these toys. Conventional laboratory-based studies examining child engagement with puzzles, parental beliefs about spatial toys, or the early development of spatial skills enable researchers to administer tightly-controlled experiments that support causal inferences, but require generalizing findings from a small subset of the population—often assuming observed effects will be invariant over chronological time, stable across different demographic groups, and consistent across distinct types of spatial play. In contrast, analyzing massive amounts of publicly available review data provides less experimental control but offers direct descriptive evidence across years, diverse populations, and variations in real-world spatial play materials without requiring statistical inference or generalization of findings from a small sample to a larger population. In the present study, we curate product reviews for spatial toys marketed to children from 0-5 years using a massive dataset of 890.7K Toys and Games product reviews collected across 8.1 million users from May 1996 to September 2023 (https://huggingface.co/datasets/McAuley-Lab/Amazon-Reviews-2023). We use structural topic modeling, an unsupervised machine learning algorithm for identifying themes in natural text data, to uncover how much people prioritize educational benefits, child enjoyment, affordability, and other factors that may drive people’s decisions to purchase spatial toys. This will enable us to examine people’s beliefs across diverse racial, socioeconomic, and educational backgrounds on a scale that would be impractical or impossible in a conventional laboratory studies. Understanding these priorities is crucial because spatial play is an important informal learning opportunity for promoting spatial skills and increasing school readiness and addressing the achievement gap in STEM. By using recent advances in Natural Language Processing to analyze authentic consumer perspectives, we can inform evidence-based policies and interventions that increase meaningful access to spatial toys, especially for families from diverse socioeconomic backgrounds. The overarching goal of this research is to bridge developmental science with the practical realities of how families make decisions about early childhood learning experiences.