Also commonly known as intelligent tutoring systems, these systems generally focus on one subject area. By focusing in one domain cognitive tutors are able to keep tabs on how the learner perceives the problem to parse misconceptions and diagnose where the learner needs help. Furthermore, these systems are good at making a student work through their problems, instead of providing the answers. Much of the cognitive tutor's ability to predict is based on the machine's neural network, similar to what you tried with Google's Quick Draw!
This is an interesting online system whereby a cognitive tutor is paired with a human tutor. The machine tutor is able to identify strengths and weaknesses, and passing that information on to a "math coach" who helps tailor the student's learning path.
This cognitive tutor boasts more than being a practice app. Instead of just repetitively completing similar math problems, students are given questions based on their competencies. MATHia provides formative assessments, support for struggling areas, and challenges when students are ready.
We have tried to focus on K-12, but found this adult intelligent tutor worth demonstrating. True to these systems, AutoTutor focuses on one subject area: reading comprehension in adults. The computer agents maintain a conversation with the user in specific knowledge domains: computer literacy, physics, critical thinking.