In some computerized educational systems, there is evidence of students "wheel-spinning," where a student tries and repeatedly fails at an educational task for learning a skill. This may be particularly concerning in low resource settings. Prior research has focused on predicting and modeling wheel-spinning, but there has been little work on how to best help students stuck in wheel-spinning. We use past student system interaction data and a minimal amount of expert input to automatically inform individualized interventions, without needing experts to label a large dataset of interventions. Our method trains a model to predict wheel-spinning and utilizes a popular tool in interpretable machine learning, Shapley values, to provide individualized credit attribution over the features of the model, including actionable features like possible gaps in prerequisites. In simulation on two different statistical student models, our approach can identify a correct intervention with over 80% accuracy before the simulated student begins the activity they will wheel spin on. In our real dataset we show initial qualitative results that our proposed interventions match what an expert would prescribe. [For the full proceedings, see ED607784.]