Developing a domain-independent measure of Computational Thinking

• In Event: Poster Session 05
  In Poster Session: PS 05 Topic - Technology, Media & Child Development

Abstract

Recent years have seen an increased emphasis on Computer Science (CS) education at the K-12 level. This not only reflects the fact that computer skills are becoming an ever-more integral part of modern life, but also results from a notion that CS education facilitates cognitive development that can be advantageous across domains. Characterized by heightened analytical thinking and problem-solving skills, this CS-supported category of cognitive development is known as computational thinking (CT). Although CT transfer remains largely untested, supporting evidence of the phenomenon would have profound impacts on intelligence theory and meaningful real-world implications for K-12 education. As a first step towards evaluating the plausibility of these claims, the present research seeks to develop and validate a domain-independent measure of CT in two different populations of children: attendees of STEM summer camps and a class of 4th graders.

Study 1. Twenty-eight children ages 7-12 (M=9.63, SD=1.21) were recruited from STEM summer camps. Some camps included CS related activities (i.e. programming and robotics) and others had non-CS focuses (i.e. chemistry and physics). Children completed a programming game (LightBot), and measures of three CS-independent cognitive CT skills: sequencing, pattern abstraction, and decomposition. The results reveal that, controlling for age, performance in LightBot (i.e. levels of progression) is highly associated with performance on the decomposition task, F(1,24)=8.72, p=0.007. This was also true of pattern abstraction, to a lesser degree, F(1,24)=6.61, p=0.017. Sequencing was found to relate to LightBot performance at only marginal significance, F(1,24)=4.12, p=0.054. These findings stand in contrast with analyses that reveal no relationship between LightBot performance and non-CT control measures of sustained attention and language comprehension (see Table 1 for summary of regression analyses).

Study 2. Thirty-one 4th grade students were recruited (17 females) from a public elementary school in which 17.70% of students are in an English Language Learner (ELL) program and 25.00% are eligible for free or reduced lunch. Most of the students of this class had little to no CS experience prior to participating. The 4th graders completed the same LightBot programming game and CT measures as the children in study 1. Results show that, of the three CT measures, only decomposition is significantly correlated with LightBot performance, r(29)=0.55, p=0.001. Perhaps as a result of the ELL demographic, English language comprehension also correlates with LightBot performance, r(29)=0.43, p=0.015 (see Table 2 for summary of correlational analyses).

Examining children from two populations with differing exposure to STEM and CS topics, these studies provide evidence for at least one cognitive, domain-independent measure of CT: decomposition skill. This demonstrates plausibility for claims that development of CT could result in the transfer of benefits to domains outside of CS. This finding paves the way for future studies to evaluate the directionality of this relationship and, ultimately, potential applications of CT skills in non-CS contexts.

Authors

• Andrew Mertens, University of Colorado Boulder
  Presenting Author

• Eliana Colunga, University of Colorado Boulder
  Non-Presenting Author