Transcript Bev-ITS
Beverly Park Woolf University of Massachusetts/Amherst U.S.A [email protected] Introduction Features of Intelligent Tutors Two Example Tutors Three Disciplines Components of Intelligent Tutors Main Drivers for a Change in Education Artificial intelligence (AI) which has led to a deeper understanding of how to represent knowledge, especially “how to” knowledge, such as procedural knowledge and reasoning about knowledge; Cognitive science has led to a deeper understanding of how people think, solve problems and learn; and The Web provides an unlimited source of information, available anytime and anyplace. • Internet provides a location--but not an education Issues addressed by this research • • • • • What is the nature of knowledge? How is knowledge represented? How can an individual student be helped to learn? What styles of teaching interactions are effective and when? What misconceptions do learners have? Intelligent Tutors Do Improve Learning • Intelligent tutors – produce the same improvement as one-on-one human tutoring and effectively reduce by one-third to one-half the time required for learning [Regian, 1997]. (One-on-one tutoring increases performance to around 98% in a standard classroom [Bloom, 1984]). – increase effectiveness by 30% as compared to traditional instruction [Fletcher, 199; Region, 1997] • Networked versions reduce the need for training support personnel by about 70% and operating costs by about 92%. • One-on-one tutoring increases performance to around the 98% in a standard classroom [Bloom, 1984]. Traditional Education Technology: • Is “frame-based” or directed; each page, every instructor response and every sequence or path of topics is predefined by the author and presented in a lock-step fashion. • Assumes that an instructional designer can specify the correct learning sequence for all students, months before a student interacts with the software. Features of Intelligent Tutors Generativity Student modeling Expert modeling Instructional modeling Self-Improving No agreement exists on which features are necessary to define an intelligent tutor. Many computer aided instructional systems contain one or more of the features listed above. Teaching systems lie along a continuum that runs from simple frame-based systems to very sophisticated and intelligent tutoring. The most sophisticated systems include, to varying degrees, these features. Features of Intelligent Tutors Generativity --(i.e., generate appropriate problems, hints and help, customized to student learning needs.) Student modeling-- (i.e. assess the current state of the student’s knowledge and learning needs and do something instructionally useful on the basis of this assessment) Expert modeling-- (i. e. assess and model expert performance in the domain and to do something instructionally useful on the basis of this assessment) Instructional modeling--(i.e., change the teaching mode based on inferences about the student’s learning). Self-Improving-- (i.e., ability to monitor, evaluate and improve its own teaching performance as a result of experience.) Assumptions of Intelligent Tutors • Intelligent reasoning can be included in educational software (e.g., simulations, games or instruction) to support both teachers and student; • Student thinking processes can be – modeled and tracked • Student actions can be predicted, understood and remediated • Teacher knowledge can be codified and carefully presented to a students AnimalWatch, Example Tutor Example of a simple addition problem in AnimalWatch AnimalWatch provided effective, confidence-enhancing arithmetic instruction for elementary students. AnimalWatch Example hint on a simple multiplication problem In contrast to common drill-and-practice systems, AnimalWatch modified its responses to conform to the students’ learning styles. The tutor presented problems that required increasingly challenging application of the cognitive subtasks involved in solving the problems (e.g. adding fractions with like denominators, adding fractions with different denominators, etc.). Cardiac Tutor 0 :3 9 Simul ati on Medic al Reco rd Chron ic Welln ess: No Signi fic ant Risk IV In Disch ar ge Comp res sing Intu ba ting Worse Not Ventil ati ng 47 / 5 Bet te r The Simulated Patient. The intravenous line has been installed (“IV in”), chest compressions are in progress, ventilation has not yet begun and the electronic shock system is discharged. The icons on the chest and near the face indicate that compressions are in progress and ventilation is not being used. Cardiac Tutor The Cardiac Tutor was generative because each case or patient situation was dynamically altered, in the middle of the case, to provide the particular arrythmia that a student needed to experience. The tutor had a complex domain model represented rules of each arrythmias and the required therapy. Nodes represented states of cardiac arrest or arrythmias and arcs represented the probability that a the simulated patient would move to a new physiological state following a specified treatment. The student model tracked student responses to each arrythmia. Student action was connected to the original simulation state so the student could request additional information about past actions. AnimalWatch AnimalWatch was generative since all math problems, hints and help were generated on the fly based on student learning needs observed by the tutor. The tutor modeled expert knowledge of arithmetic as a topic network with nodes such as ``subtract fractions'' or ``multiply whole numbers.” Student modeling dynamically recorded each sub-task learned or needed based on student action. The tutor was self-improving in that it used machine-learning techniques to predict how long a student needed to solve a problem and each student’s proficiency. This Research Area Encompasses Several Disciplines Tools and Methods are derived from: • Artificial Intelligence – Design and build systems that exhibit intelligence • Cognitive Science – Investigates how intelligent entities (human or computer) interact with their environment, and acquire • Education – Explore effective methods of supporting teaching and learning New Disciplines Have Formed Represent Domain Knowledge Fractions Fractions Fraction Readiness Addition /Subtraction Similar Denominators Multiplication/Division Dissimilar Denominators Find LCM Multiply All fractions Represent Domain Knowledge Cardiac Resuscitation •vtach •vtach •vfib •asys •brady •vfib •25% •asys •brady •GOAL •sinus