Computational models of cognitive development could be used to optimize educational interventions by predicting their outcomes. Similar to how computational screening of drug candidates guides clinical trials in medicine, models could screen millions of potential educational interventions to identify the most promising ones for empirical testing. This would help address the challenge of limited resources to test all possible interventions. Open questions remain around how accurately models can predict real-world results and whether connectionist models are sufficiently grounded in biology for this application.