Autotutor combines a talking head with research in discourse and tutoring dialogs. The most novel is the use of Latent Semantic Analysis (LSA), a statistical technique that can estimate the similarity between two pieces of text. By developing databases of good and bad expected answers, and comparing the similarity of student responses to these expected responses, the tutoring system can produce a type of dialog of human tutors in guiding and correcting students in learning. One of the keys is the system capability to provide the learner with a personalized, adaptive effective teaching. In order to develop such an agent, Autotutor trained a feed-forward, backpropagation neural network to predict the number of errors a student will make. The achieved prediction accuracy is high, showing that a neural network is capable of making such predictions.