Academic journal article Educational Technology & Society

Passively Classifying Student Mood and Performance within Intelligent Tutors

Academic journal article Educational Technology & Society

Passively Classifying Student Mood and Performance within Intelligent Tutors

Article excerpt


The goal of this research was to develop an adaptable ITS conceptual model that includes appropriate inputs to determine the affective state of the student being tutored. In support of this goal, we surveyed methods to allow an ITS to classify the affective state (e.g., mood or emotional state) of the student through the passive sensing and interpretation of correlated student behaviors and their physiological responses during training. We investigated methods of classifying affect for virtual characters through procedural reasoning systems and adapted these methods to human students to classify the student's affective state. Understanding the student's affective state, the ITS can use that information along with other student data (e.g., knowledge and progress toward goals) to select instructional strategies (e.g., direction) to optimize learning and performance.

The research discussed below offers perspectives on: one-to-one tutoring; affect and learning; the need for ITS to be capable of "perceiving" student affect; the design limitations of current ITS; models of affect; and enhanced ITS models. The main body of our research contains two primary objectives: the review of methods to sense student behaviors unobtrusively (passively) so as avoid any negative impact on the learning process; and the interpretation of sensed behaviors to build a predictive model of student affect as a basis for making decisions on the delivery of instruction.

A perspective on one-to-one tutoring

An ongoing goal in the research and development of ITS has been to increase their adaptability to better serve student needs (Heylen, Nijholt, op den Akker & Vissers, 2003; Hernandez, Noguez, Sucar & Arroyo-Figueroa 2006 and Sottilare, 2009). "The basic tenet of intelligent tutors is that information about the user (e.g., knowledge, skill level, personality traits, mood level or motivational level) can be used to modify the presentation of information so that learning proceeds more efficiently." (Johnson & Taatgen, 2005, p. 24).

ITS offer the advantage of one-to-one tutoring where instruction is delivered at a tailored pace based on competency and progress toward instructional goals. The value of expert, one-to-one, human tutoring vice group tutoring (i.e. traditional classroom teaching) has been documented among students who often score 2.0 standard deviations higher than students in a conventional classroom (Bloom, 1984). Other advantages of one-to-one, human-tutoring over classroom settings is that students ask approximately 26.0 questions per hour versus less than 0.20 questions per hour in the classroom setting (Dillon, 1988; Graesser & Person, 1994). This higher rate of interaction provides additional learning opportunities for weaker students. Stronger students ask fewer questions, but these questions tend to be "deep-reasoning questions" (Person & Graesser, 2003).

Loftin, Mastaglio & Kenney (2004) assert that while one-to-one human tutoring is still superior to ITS in general, the one-to-one human tutoring approach is neither efficient nor cost-effective for training large, geographically- distributed populations (e.g., military organizations or large multi-national corporations). Given large and potentially diverse student populations, one goal of our research was to identify methods to improve the adaptability of ITS to support one-to-one tutoring. We explored methods to increase ITS "perception" of the student's affective state and thereby theoretically increase the potential of the ITS to effectively adapt to a given student during one-to-one, computer-based instruction.

A perspective on affect and learning

Linnenbrink and Pintrich (2002) identified a connection between affect and learning. They found that many students experience some confusion when confronted with information that does not fit their current knowledge base, but those in a generally positive affective state adapted their known concepts to assimilate the new information. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.