This study investigates the role of grit, a non-cognitive skill indicating perseverance and passion for long-term goals, in predicting academic resilience during the COVID-19 pandemic. Using data from a digital learning platform in the United Arab Emirates, the researchers construct a behavioral measure of grit based on students' responses to performance setbacks. They find that students with higher grit levels before the pandemic experienced smaller declines in math and science scores during the pandemic. Machine learning analysis shows that behavioral data from the platform can explain 77% of the variance in academic resilience, outperforming survey-based measures of grit. The study highlights the importance of non-cognitive skills in educational outcomes and suggests that digital learning platforms can provide valuable insights into student behavior and performance. The findings have implications for interventions aimed at fostering grit and improving educational outcomes, especially with the increasing use of AI-based learning technologies. The study also notes that behavioral data from digital platforms can help identify students who may struggle with resilience or motivation, enabling targeted interventions. The research underscores the potential of digital learning platforms not only for delivering education but also for assessing and fostering non-cognitive skills. The study's results suggest that behavioral measures of grit are more reliable than self-reported surveys in predicting academic resilience during the pandemic. The study also acknowledges limitations, including potential biases in survey data and the possibility that results may be country or age-specific. Overall, the study provides evidence that grit is a critical non-cognitive skill that can predict academic performance and resilience, especially in challenging circumstances like the pandemic.This study investigates the role of grit, a non-cognitive skill indicating perseverance and passion for long-term goals, in predicting academic resilience during the COVID-19 pandemic. Using data from a digital learning platform in the United Arab Emirates, the researchers construct a behavioral measure of grit based on students' responses to performance setbacks. They find that students with higher grit levels before the pandemic experienced smaller declines in math and science scores during the pandemic. Machine learning analysis shows that behavioral data from the platform can explain 77% of the variance in academic resilience, outperforming survey-based measures of grit. The study highlights the importance of non-cognitive skills in educational outcomes and suggests that digital learning platforms can provide valuable insights into student behavior and performance. The findings have implications for interventions aimed at fostering grit and improving educational outcomes, especially with the increasing use of AI-based learning technologies. The study also notes that behavioral data from digital platforms can help identify students who may struggle with resilience or motivation, enabling targeted interventions. The research underscores the potential of digital learning platforms not only for delivering education but also for assessing and fostering non-cognitive skills. The study's results suggest that behavioral measures of grit are more reliable than self-reported surveys in predicting academic resilience during the pandemic. The study also acknowledges limitations, including potential biases in survey data and the possibility that results may be country or age-specific. Overall, the study provides evidence that grit is a critical non-cognitive skill that can predict academic performance and resilience, especially in challenging circumstances like the pandemic.