Artificial intelligence and computer science have advanced to the point that computer systems that are as successful as smart human tutors are feasible. Intelligent tutoring computer systems are being built to provide the same educational benefit that a knowledgeable human tutor can provide to students. A good teacher in private knows the student and listens to the special needs of the student. The machine has been seen from the beginning as offering such instruction, thereby having the potential to improve educational quality. Experience has shown that the learning process is highly effective for students seeking professional, in-person tutoring. It is a key self-regulatory ability to know when and how to seek help during learning. Classroom research suggests that adaptive behavior in help-seeking helps students learn more effectively. Help requests (and responses) can be characterized as “instrumental” when aimed at learning, or “executive” when aimed solely at tasks to be completed. Active help-seeking in educational technology has also been shown to be correlated with better learning. A more comprehensive solution would be to help students gain a better ability to seek support.
Intelligent tutoring programs are attempting to emulate best practices developed from the human tutoring model. For instance, the smart tutor will usually assess each step of the learning process of the students and provide the student with tips and suggestions to figure out the problem on their own. Intelligent systems are distinguished from other computer-aided instruction in that, by building a student learner model, they can interpret and respond to student interactions in real-time. Intelligent tutoring programs are trying to figure out not only whether or not an answer is wrong, but also how the student has achieved that understanding. To design efficient programs, intelligent tutors rely heavily on cognitive theories and practices. Some of the hypotheses of cognitive science and educational psychology involve reducing a load of working memory by supplying the learner with a “scratchpad” to work on. Additionally, the tutors typically utilize problem-solving behaviors that are goal-driven. It has been shown that intelligent tutoring systems (ITS) are particularly valuable for teaching cognitive tasks such as troubleshooting, problem-solving, and critical situations. Like a human teacher, an ITS constantly tracks and analyses the behavior of the individual student, infers the state of knowledge of the student, and decides on the next teaching step to optimize the learning of the student based on an embedded student model, assignment model, and instructional model.
Robust learning of help-seeking skills would enable students to transfer these to novel learning skills situations when there is no explicit support for help-seeking. The effect of feedback on the learning of students is understood. Metacognitive feedback is defined as feedback triggered by the learning behavior of students (e.g. avoiding needed help) and not by the accuracy of their domain-level responses. Meta-cognitive feedback also provides metacognitive content, i.e. it transmits information on desired learning behaviors (e.g., advising the student to ask for a hint) rather than domain knowledge. A small number of usemetacognitive feedback systems in their learning process to direct students. However, the effect of metacognitive feedback on the help-seeking behavior of students is still to be fully evaluated in controlled studies to the best of our knowledge.