Exploring cognitive diversity in a sample of children from Unsatisfied Basic Needs homes

Child diversity has become an increasingly relevant topic in developmental science with significant methodological and applied implications. Clustering algorithms, a type of machine learning approach, offer accessible methods for exploring such diversity by identifying the latent data structure, which allows for the extraction of groups or profiles of subjects. Following a person-oriented approach, this study explores the cognitive profiles of a sample of 105 children coming from Unsatisfied Basic Needs homes in Argentina. An ensemble clustering method was applied to performance data from tests assessing attention, inhibitory control, working memory, planning, and fluid reasoning. Results delivered three specific clusters with statistically significant differences in cognitive accuracy and reaction time variables. Specifically, we found a "working memory low-performance group," a "generalized high-performance group," and a "generalized low-performance group." The individual characteristics of these clusters may potentially guide differentiated strategies for future interventions. In addition, the incorporation of diverse performance variables - such as accuracy and response times - can help uncover individual strengths and weaknesses that may inform the design of such interventions. This approach contributes to fostering more inclusive and diverse perspectives in developmental science, moving away from deficit-based interpretations of low or atypical performance and instead recognizing them as expressions of the potential diversity of developmental pathways.