Decoding agent-based models supports students’ mechanistic and causal reasoning about scientific phenomena

A common rationale for integrating computational thinking (CT) in science curricula has been the opportunity to increase learning outcomes in both CT and science. While evidence shows that learning to code to create computer models of scientific phenomena improves students’ CT, few studies have demonstrated equivalent increases in science learning. This study aims to investigate the impact of a CT integration curriculum featuring “decoding”, or explicitly mapping between mechanisms in code and processes in science, on science learning, specifically about ecosystems, and CT across three cohorts of middle school students focusing on their survey scores and artifact-based interviews. Our research involved 70 middle school students, 38 of whom received the intervention and 32 of whom were in the control. In this mixed-methods study, quantitative analysis of pre-, post- and continuation survey data was used to measure the impact of decoding and case studies were used to elucidate how students’ science and CT learning were impacted by decoding. We found that the treatment group students from all three cohort years (n = 46) on average significantly improved their KSCT scores after the workshop with an effect size in the high range (Cohen’s d = .96). Through the analysis of rich and thick qualitative data collected through artifact-based interviews, we identified how decoding skills impacted students’ reasoning about scientific phenomena, and how students used their decoding skills to make sense of a new scientific phenomenon. This study adds to the literature on CT integration in science education by elucidating an approach to deepen science learning and CT synergistically through the study of mechanisms. The findings from this study indicate that CT integration using the Decoding Approach yielded student gains in science learning and CT. Our research also advances knowledge of how and why CT integration, specifically through decoding, can support science learning and CT. Evidence from four qualitative cases that shows how students use coded mechanisms as an “active” and “executable” representation with which to reason about scientific phenomena while decoding. The research highlights that students do not need to make computer models from scratch to benefit from CT integration. We show that decoding models made by others is a valuable learning opportunity that impacts students’ MR, decoding skills, and science learning. The results from this study imply there are untapped opportunities for students to gain an understanding of ecosystems, and in some cases transfer that understanding to new systems, through the integration of CT with a particular emphasis on decoding.