Personalized learning considers that the causal effects of a studied learning intervention may differ for the individual student. Making the inference about causal effects of studies interventions is a central problem. In this paper we propose the Residual Counterfactual Networks (RCN) for answering counterfactual inference questions, such as “Would this particular student benefit more from the video hint or the text hint when the student cannot solve a problem?”. The model learns a balancing representation of students by minimizing the distance between the distributions of the control and the treated populations, and then uses a residual block to estimate the individual treatment effect based on the representation of the student. We run experiments on semi-simulated dataset and real-world educational online experiment dataset to evaluate the efficacy of our model. The results show that our model matches or outperforms the state-of-the-art.