Markus Thiel

Opaque computational simulations pose a distinctive epistemic challenge for learners: they must evaluate outputs generated by processes they cannot directly inspect. How students navigate this condition in situated discourse remains insufficiently understood. This study examined this challenge in the context of molecular dynamics (MD) simulations, where well-established physical principles are implemented through computational procedures that remain inaccessible to students. Eight undergraduate chemistry students completed a five-hour Jupyter Notebook-based laboratory session on MD simulations and subsequently participated in three semi-structured focus group interviews. The interviews were analyzed through structured qualitative content analysis, guided by the Epistemic Practices in Opaque Computational Learning Environments – Chemistry and Modeling (EPOCLE-CM) analytical framework, yielding 189 coded segments across five epistemic dimensions and a supplementary Domain Knowledge category. Three configurations of epistemic practice emerged: Reflective Model Evaluation, Boundary-Based Model Critique, and Calibrated Trust. Trust did not function as a stable attitude but was produced through sustained reasoning about the simulation–experiment relationship: students calibrated their reliance on outputs by reflecting on representational scope, model boundaries, and conditions under which simulation results could count as evidence (φ = .509 between trust calibration and simulation–experiment reflection). For instruction, boundary testing and deliberate use of simulation breakdowns emerged as productive entry points for fostering epistemic engagement with opaque computational tools. As computational systems become less transparent, understanding how learners navigate opacity is relevant beyond chemistry education.

Scientific reasoning is a multidimensional construct; therefore, assessment claims should address specific processes elicited by an instrument. This study presents a game-based assessment (GBA) that uses log-derived decision outcomes to examine planning investigations as a process-related dimension of scientific reasoning among 47 preservice chemistry teachers. In the assessment, planning investigations were operationalized as determining an appropriate analytical procedure, represented by the selected sample, analytical method(s), and procedural order. Across three region-specific tasks, within an environmental chemistry scenario, participants identified the sample relevant to the problem, selected up to two analytical methods, and sequenced them. According to the findings, the scores differed significantly across the three tasks with the highest performance in the rocky area and the lowest performance in the forest area task. These findings highlight the importance of examining planning decisions across multiple tasks. Thus, the game offers a way to assess these planning investigations through gameplay records.