ABSTRACT Scientific modeling is a core practice of scientific inquiry. Students' engagement in modeling can be enhanced by attending to epistemic criteria, which in science are standards used to evaluate the validity and accuracy of scientific models. While prior research has focused on students' development and use of epistemic criteria in scientific inquiry environments, less is known about students' metacognitive knowledge about why epistemic criteria matter—their metacognitive epistemic justifications. Investigating students' metacognitive knowledge of criteria can reveal their understandings of the criteria guiding scientific practice (such as modeling), their relative value, and their relationship to epistemic products. We present findings from clinical interviews of student dyads conducted at the end of the implementation of a fifth‐grade model‐based inquiry unit in which the class developed and used epistemic criteria for scientific models. Students' justifications of epistemic criteria revealed their metacognitive knowledge about why criteria, such as evidentiary fit, are important in modeling practice. We also identified themes in students' justifications for the relative importance of different epistemic criteria and the role such justifications played in resolving disagreements about epistemic matters. Our findings illustrate the potential value of engaging students in discourse about metacognitive justifications for epistemic criteria.

Scientific modeling is supported by epistemic criteria to evaluate the "goodness" of scientific products.A central epistemic criterion is causal mechanism which means identifying the underlying causes which generate changes within a system.During a five-week model-based inquiry unit students identified causes of the phenomenon, eutrophication, but it was unclear whether student understood the causes hierarchically, i.e., whether some causes led to others, and whether some were more critical to include in a model than others.In a model evaluation task, students prioritized a model that was complete and had all the causes, deprioritized the model missing the direct causal mechanism, and placed models missing the initial causes in the middle.The results indicate that students thought about mechanisms in terms of chaining which means students thought about how causes led to outcomes within a system.

Reforms in science education have called for students to engage in scientific practices. A scientific practice that has received increased attention is modeling. Part of the practice of modeling is the development, use, and evaluation of epistemic criteria which are the standards scientists use to evaluate the validity and accuracy of scientific models. For students to effectively engage with epistemic criteria, they need to be given collective cognitive responsibility, or epistemic agency. This study investigates how the design and implementation of a fifth-grade model-based inquiry unit supported students' epistemic agency to revise an initial class criteria list for scientific models. The findings indicate that reflecting upon peer critique of their scientific models in a written task, and the ways that teachers engaged students in a class discussion to revise the initial class list, supported students' epistemic agency to revise an initial class criteria list for scientific models.

A promising way to mitigate inequality is by addressing students' worries about belonging. But where and with whom is this social-belonging intervention effective? Here we report a team-science randomized controlled experiment with 26,911 students at 22 diverse institutions. Results showed that the social-belonging intervention, administered online before college (in under 30 minutes), increased the rate at which students completed the first year as full-time students, especially among students in groups that had historically progressed at lower rates. The college context also mattered: The intervention was effective only when students' groups were afforded opportunities to belong. This study develops methods for understanding how student identities and contexts interact with interventions. It also shows that a low-cost, scalable intervention generalizes its effects to 749 4-year institutions in the United States.