Measuring student mindsets at scale in resource-constrained settings: A toolkit with an application to Brazil during the pandemic.

Abstract

Mounting evidence that growth mindset-the belief that intelligence is not fixed and can be developed-improves educational outcomes has spurred additional interest in how to measure and promote it in other contexts. Most of this research, however, focuses on high-income countries, where the most common protocols for measuring and intervening on student mindsets rely on connected devices-often unavailable in low- and middle-income countries' schools. This paper develops a toolkit to measure student mindsets in resource-constrained settings, specifically in the context of Brazilian secondary public schools. Concretely, we convert the computer-based survey instruments into text messages (SMS). Collecting mindset survey data from 3570 students in São Paulo State as schools gradually reopened in early 2021, we validate our methodology by matching key patterns in our data to previous findings in the literature. We also train a machine learning model on our data and show that it can (1) accurately classify students' SMS responses, (2) accurately classify student mindsets even based on text written in other media, and (3) rate the fidelity of different interventions to the published growth mindset curricula.