Intelligent Tutoring Systems

Intelligent Tutoring Systems (ITS) are computer-based training programs that use artificial intelligence to tailor multimedia learning by providing individualized instruction (Ong & Ramachandran, 2000). Intelligent Tutoring Systems try to imitate the help that a live tutor would provide to an individual student (Johnson, 2005). ITS offer a way to identify, remediate, and track all students separately (Johnson, 2005). The goal of ITS is to provide the benefits of individualized instruction without the cost and time it takes to provide personalized instruction with teachers (Ong & Ramachandran, 2000).

Keywords Acquisition; Computer Aided Instruction (CAI); Computer-Based Training (CBT); Distance Learning; Individualized Instruction; Instructional Technology; Intelligent Tutoring Systems (ITS); Module; No Child Left Behind Act of 2001(NCLB); Retention

Overview

The concept of intelligent tutoring systems (ITS) has been studied for more than thirty years by researchers in education, psychology, and artificial intelligence (Ong & Ramachandran, 2000). Schools and universities have been looking for ways to increase learning for students and improve test scores (Wijekumar, 2006). Since computer technologies have made their ways into both homes and schools, computer tutoring systems are viewed as a potential solution to this problem (Wijekumar, 2006). In the early 1990s, Jay Liebowitz wrote a book called The Explosion of Intelligent Systems by the Year 2000 in which he made some apt predictions about the future of technology (cited in Karlin, 2007). He imagined a world in which students of all grade levels and abilities used video-conferencing in schools, employed voice-activated programs, and enrolled in distance learning courses (cited in Karlin, 2007). It turns out he was right. By the early years of the twenty-first century, ITS were providing interactive instruction to support K-12 education, college education, corporate training and military preparation.

What are Intelligent Tutoring Systems?

Intelligent Tutoring Systems (ITS) are computer-based training programs that use artificial intelligence to tailor multimedia learning by providing individualized instruction (Ong & Ramachandran, 2000). They are also referred to as intelligent computer-aided instruction (ICAI) and have been a major breakthrough in the field of instructional technology. Before, ITS, computer-based training (CBT) and computer aided instruction (CAI) were the only computer teaching systems (Beck, Stern, & Haugsjaa, 2004). In systems like these, the directions were not specified to meet the individual needs of each learner, and transitioning a student through the material was formulaic and inflexible, “such as ‘if question 21 is answered correctly, proceed to question 54; otherwise go to question 32’” (Beck, Stern, & Haugsjaa, 2004, ¶ 1). Prior knowledge and learning style were not taken into account. As a result, their impact on learning was mediocre (Wijekumar, 2006).

Intelligent Tutoring Systems are more advanced, allowing learners to hone their abilities by completing assignments within interactive academic settings. ITS can answer questions and provide personalized assistance to the learner. ITS, unlike other educational technologies, evaluate every student’s response in order to assess his/her knowledge and skills (Ong & Ramachandran, 2000). ITS can then modify instructional strategies, give explanations, examples, demonstrations, and practice exercises where necessary (Ong & Ramachandran, 2000). ITS offer more options in the presentation of material and have the capability to specialize information to cater to a student's needs (Beck, Stern, & Haugsjaa, 2004).

The typical ITS model does the following:

• Identifies learning objectives and their context;

• Acknowledges gaps in individual student's knowledge;

• Trains each student according to the areas in which they lack knowledge;

• Guides the student through the relevant parts of the book, or provided material;

• Assesses students on the learning objectives;

• Gives the student feedback on his/her responses and provides explanation as to why an answer is correct or incorrect; and

• Provides each student with more questions in the specific areas where they lack knowledge (Sessink, Beeftink, Tramper, & Hartog, 2007).

The inception of the No Child Left Behind Act has put pressure on schools to deliver high-quality instruction to all their students (Wijekumar, 2006). As a result, schools are trying to utilize technological advancements such as Intelligent Tutoring Systems to teach their subjects, practice tests, and track progress (Wijekumar, 2006). Intelligent Tutoring Systems provide motivation, modeling, interactivity, feedback, and consistency like no other tool before (Wijekumar, 2006).

Why Use Intelligent Tutoring Systems?

Many academic courses are attended by a heterogeneous group of students who have come from different backgrounds and have attended different courses in the past; some even speak in different languages. Some students may simply be at varying skill levels. (Sessink, Beeftink, Tramper, & Hartog, 2007). Effectively teaching a heterogeneous student population is a challenge in education because most traditional methods target the average student. This is a definite disadvantage for advanced students, students with disabilities, and students who lack certain prior knowledge (Sessink, Beeftink, Tramper, & Hartog, 2007). Unfortunately, class sizes and instructor loads often make it impossible for teachers to tutor every student individually. Intelligent Tutoring Systems try to imitate the help that a live tutor would provide to an individual student (Johnson, 2005). ITS offer a way to identify, remediate, and track all students separately (Johnson, 2005).

The main initiative of ITS is giving personalized instruction without the cost and time it takes to provide personalized instruction with teachers (Ong & Ramachandran, 2000). ITS can be thought of as a virtual training assistant that collects the wisdom and experience of trained teaching professionals and distributes the content to students electronically (Ong & Ramachandran, 2000). Adaptive Intelligent Tutoring systems may help to support learning for a heterogeneous group of students.

Research on prototype Intelligent Tutoring Systems indicated that students who used ITS generally learned faster and demonstrated improved performance compared to classroom-trained participants. In the 1980s, Benjamin Bloom determined “that students who receive one-on-one instruction perform two standard deviations better than students who receive traditional classroom instruction. An improvement of two standard deviations means that the student performed in the top 2 percent of those receiving instruction” (Ong & Ramachandran, 2000, ¶ 18).

In a more recent study conducted at Carnegie Mellon University, college students used an ITS called the LISP Tutor to learn computer programming skills (Ong & Ramachandran, 2000). “Students who used the ITS scored 43 percent higher on the final exam than the control group” (Ong & Ramachandran, 2000, ¶ 4). In addition, the control group needed 30 percent more time to solve complex programming problems (Ong & Ramachandran, 2000). In another example, students using Smithtown, an ITS for economics, did not perform better than students in a traditional learning environment, but they required less time to cover the material (Beck, Stern, & Haugsjaa, 2004). Performance was equal but more efficient.

In 2011 VanLehn compared computer tutors and human tutors for their impact on learning gains, particularly focusing on experiments that compared one type of tutoring to another while attempting to control all other variables, such as the content and duration of the instruction (VanLehn, 2011). He concluded that “ITS should be used to replace homework, seatwork, and perhaps other activities but not to replace a whole classroom experience. Nonetheless, within their limited area of expertise, currently available ITS seem to be just as good as human tutors” (VanLehn, 2011, p. 214).

Further Insights

How Intelligent Tutoring Systems Work

Many traditional teaching methods introduce learners to facts and concepts and follow up with test questions to assess understanding (Ong & Ramachandran, 2000). “These methods are effective in exposing people to large amounts of information and testing their recall but learners often” are not taught how to correctly apply their new knowledge (Ong & Ramachandran, 2000, ¶ 5). By contrast, Intelligent Tutoring Systems use highly interactive learning environments, including simulations, that require students to use the skills they have just acquired. This type of learning is effective in helping students retain and apply knowledge more effectively in the future (Ong & Ramachandran, 2000).

In order to provide such specialized guidance to students, ITS systems typically use three kinds of knowledge, arranged into the following software components:

• The expert model (also known at the domain model),

• The student model and

• The instructor model (also known as the adaptive or pedagogical model).

”The expert model represents subject matter expertise and provides the ITS with knowledge of what it's teaching” (Ong & Ramachandran, 2000, ¶ 7). The student model shows the student what he or she knows as well as what he or she has yet to learn, letting the system know what type of learner it is teaching. The instructor model uses information from the other two models and assembles appropriate content and provides the necessary instructional strategies through the ITS user interface (Sessink, Beeftink, Tramper, & Hartog, 2007).

The Expert Model

The expert model is the backbone of the ITS structure. It possesses the content that the instructor is teaching (Beck, Stern, & Haugsjaa, 2004). The knowledge in the expert model provides the system with information with which to compare the learner's selections so that the ITS can evaluate the learner's previous knowledge as well as the knowledge gained through the use of the system (Ong & Ramachandran, 2000). For that purpose, the expert model is more than just a representation of data; it is a prototype of how someone skilled in a particular subject area applies knowledge (Beck, Stern, & Haugsjaa, 2004). This allows the ITS to more accurately pinpoint a learner's problem areas.

The Student Model

The student model “evaluates student performance to determine his or her knowledge and skills” (Ong & Ramachandran, 2000, ¶ 10). By maintaining a record of each user's skills and drawbacks, the ITS can provide effective, individualized instruction (Ong & Ramachandran, 2000). The student model keeps its individualized content in its electronic storage, allowing for easy access to each user (Beck, Stern, & Haugsjaa, 2004). The information gathered shows what the system sees as the learner's current skill level (Beck, Stern, & Haugsjaa, 2004). In order for the learner to achieve training, the student model should improve for each learning objective with which the user is presented and after a user completes a module, the student model should be on a satisfactory level for each learning objective (Sessink, Beeftink, Tramper, & Hartog, 2007).

A student model should contain a record of the student's understanding of the material, as well as more general information about the student such as learning preferences, acquisition and retention (Beck, Stern, & Haugsjaa, 2004). Acquisition calculates the time it takes for students to learn new material. Retention evaluates the recall of the learning content over time. Studies have shown that “examining a learner's acquisition and retention levels can be beneficial for modeling instruction. At a minimum, such a model traces how well a student is performing on the material being taught” but it may also record the things a student does not understand (Beck, Stern, & Haugsjaa, 2004, ¶ 31). A more intuitive ITS, such as the one used in Naval Officer training, can monitor student’s electronic actions to determine where his or her level of comprehension lies (Ong & Ramachandran, 2000). That particular system uses “pattern matching to detect sequences of actions that track whether the student does or doesn't understand” (Ong & Ramachandran, 2000, ¶ 13).

In their study of methods for constructing more accurate student models, Gong, Beck, and Heffernan conclude, “one commonality across all of the models tested is they are much more accurate in making predictions when students respond to items correctly than when students respond incorrectly. Understanding the reason for this phenomenon, and developing models that do a better job at handling these cases, is a major open challenge for the student modelling community.” (Gong, Beck, & Heffernan, 2011, p. 43).

The Instructor Model

The instructor model provides the instructional strategies that are determined to be most effective for teaching the assigned material to “the learner based on its knowledge of the user's strengths and weaknesses and preferred learning styles” (Ong & Ramachandran, 2000, ¶ 14). One concern for the instructor model of “an ITS is the selection of a meta strategy for teaching the domain. For example, the system could decide to use the Socratic method or it could select a series of problem-solving examples” (Beck, Stern, & Haugsjaa, 2004, ¶ 32).

This component uses data from the student model to provide an appropriate model for teaching the specific user (Beck, Stern, & Haugsjaa, 2004). Types of information that are regulated by the instructor model are when to review, “when to present a new topic, and which topic to present” (Beck, Stern, & Haugsjaa, 2004, ¶ 7). If a student was determined to be a novice in a certain area of instruction, then this model might include a review or show a step-by-step demonstration before asking the student to apply his or her own knowledge (Ong & Ramachandran, 2000).

The instructor model may also give constructive feedback and detailed coaching while the learner complete relevant exercises (Ong & Ramachandran, 2000). As a learner develops his/her skills, the instructor model may scale back on hints and examples and present increase the complexity of the exercises (Ong & Ramachandran, 2000).

Types of ITS Instruction

Recent work in organizing information has been derived from the idea that instructional goals should vary according to the type of information learned. In addition to this goals should be set with regards to what the student will be able to do when they complete the ITS lesson (Beck, Stern, & Haugsjaa, 2004).

Most Intelligent Tutoring Systems tend to focus on teaching procedural skills. The instructional aim is for learners to complete a certain task (Beck, Stern, & Haugsjaa, 2004). Systems designed with this objective are often called cognitive tutors. They are usually structured around a set of guidelines that are part of the expert model. This set of expert rules also can serve as the knowledge domain for the instructor model. If a student approaches difficult or trying questions, the expert model can determine the most effective way to assist (Beck, Stern, & Haugsjaa, 2004). Many systems try to teach procedures by taking on the appearance and manner of a real educational environment (Beck, Stern, & Haugsjaa, 2004). A simulated learning environment can help improve the benefits of educational training and diminish the costs (Beck, Stern, & Haugsjaa, 2004).

Contrary to the simulation based tutors are the teachers of information that lies outside of the real world context (Beck, Stern, & Haugsjaa, 2004). These Intelligent Tutoring Systems focus on teaching concepts. These systems are referred to as knowledge based tutors and require more substantial domain knowledge. They promote complex problems that the student can solve without with out requiring a connection to real life issues. ITS of this type use general teaching strategies and emphasize the explication and presentation system to accomplish educational objectives. They are arranged in order to provide students with abstract knowledge that can correlate with other, future situations (Beck, Stern, & Haugsjaa, 2004).

Viewpoints

Challenges Associated with Intelligent Tutoring Systems

Difficult to Program

Intelligent tutors are available but not used as much as Liebowitz had predicted in The Explosion of Intelligent Systems by the Year 2000 due to the labor-intensive programming required to effectively and intuitively deliver course material (Karlin, 2007). Development of ITS are not only time consuming but require specific knowledge and advanced computer skills (Sessink, Beeftink, Tramper, &Hartog, 2007). “Developing an expert system that provides comprehensive coverage of the subject material is difficult and expensive” (Ong & Ramachandran, 2000, ¶ 9). This can take months or years and often a design team is necessary to create a successful system (Montalvo, 2006).

Ansari and Sykes (2012) proposed the use of enthymemes, a manner of presenting a deductive argument, in programming tutors and suggest that the framework of an ITS could benefit greatly by incorporating enthymemes in the transfer of knowledge to the interactive users of such systems. The study authors assert that “by including enthymemes and transformation rules, Intelligent Tutoring Systems may in the near future offer what they have always intended to do--provide personalized instruction comparable to a domain expert tutor” (Ansari & Sykes, 2012, p. 28).

Limited User Communication

Another challenge with ITS is that is limited in terms of communication with the user (Beck, Stern, & Haugsjaa, 2004). Therefore it is difficult to get an accurate representation of learner's abilities and progress (Beck, Stern, & Haugsjaa, 2004). As a result, the student model of the ITS may not be accurate. An ITS that has misread a learner's abilities may offer too much assistance to a learner who is performing satisfactorily and vice versa. Yet, even an imperfect student model can be valuable in tailoring education methods to a student's individual abilities. This is what happened with earlier CBT and CAI systems, which could not individualize instruction (Beck, Stern, & Haugsjaa, 2004).

Delays in Technical Assistance

One of the biggest challenges with using ITS is responding to the user's request for assistance. “When faced with a problem, the student must wait for help from technical support or an online instructor who may take a long time to respond” (Ong & Ramachandran, 2000, ¶ 7).

Other Variables

While the potential for using ITS in schools and other learning environments continues to grow, there are limits to what can be achieved (Wijekumar, 2006). There can be inconsistencies in operating systems, technology support, and computer literacy among students (Wijekumar, 2006). In addition, sociological, political, and economic variables influence the effect of ITS on the learning environment (Wijekumar, 2006). All of these factors can effect the success or failure rates of ITS in schools.

Potential Solutions

A common alternative to the expensive process of customizing ITS is to allow instructors or schools to supply much of the knowledge needed to achieve learning goals. This method avoids having to cover all possible problems in an ITS (Ong & Ramachandran, 2000). Rather, it needs only a way to indicate that what the learner needs to know is corresponding with how to apply the knowledge in real setting (Ong & Ramachandran, 2000).

Authoring Tools

Authoring tools are a great technology to help make customing ITS easier and more efficient. The goal of authoring tools is to allows educators to use a basic development shell to author their own course material within an Intelligent Tutoring System and to allow programmers to input more to the expert, student and instructor models. (Beck, Stern, & Haugsjaa, 2004). Effective authoring tools could empower teachers to have more control over the material presented and would allow them to quickly modify and update material as needed. It would also allow fewer developers to design software (Beck, Stern, & Haugsjaa, 2004).

Although the planning and implementation phase of ITS development can be time-consuming, research has shown that, once designed, Intelligent Tutoring Systems can help minimize an instructor's workload compared to face-to-face instruction (Montalvo, 2006).

Conclusion

Technologies for learning environments are a vital tool for the ever-changing field of education. Computer-based tutoring programs have been designed to model direct instruction while using the technical ability to provide effective feedback, remediation, and guided practice to all learners (Magliaro, Lockee, & Burton, 2005). Intelligent tutoring systems have proven their extreme efficiency and aid in increasing student education. In addition to developing and enhancing systems, an important consideration in the advancement of ITS is reducing development time and expense (Beck, Stern, & Haugsjaa, 2004). Solving this problem would allow more systems to be created and could redirect research foci on how to make ITS more effective. The current emphasis on standardized testing and accountability has created opportunities for ITS to enhance traditional instruction and maximize both the learner and the educator's time (Magliaro, Lockee, & Burton, 2005).

Terms & Concepts

Acquisition: Acquisition refers to how quickly students learn new material.

Computer Aided Instruction (CAI): Less sophisticated than Intelligent Tutoring Systems, CAI refers to computer technology that assists the teaching and learning process.

Computer-Based Training (CBT): Like CAI, CBT refers to instruction and learning that takes place using computer technology.

Distance Learning: The term distance learning is applied to instruction that occurs remotely, rather than in person. Teachers and students may communicate via mail, telephone, email or online programs.

Individualized Instruction: Individualized Instruction is the term applied to teaching practice that considers the needs of each individual student and customizes instruction accordingly.

Instructional Technology: The theory and practice of designing, developing, managing and evaluating processes and resources for learning (Moore, 2006).

Intelligent Tutoring Systems (ITS): Intelligent Tutoring Systems (ITS) are computer-based training programs that use artificial intelligence to tailor multimedia learning by providing individualized instruction (Ong & Ramachandran, 2000).

Module: This term refers to a collection of learning objectives and corresponding learning material within a set ITS program or course (Sessink, Beeftink, Tramper, & Hartog, 2007).

No Child Left Behind Act of 2001(NCLB): A United States federal law that aims to streamline education and narrow the achievement gap across the nation's schools.

Retention: Retention is the term used to describe a person's recall of material over time.

Bibliography

Ansari, S., & Sykes, E. R. (2012). Towards smarter intelligent tutoring systems: A proposal for the inclusion of enthymemes in their design. Technology, Instruction, Cognition & Learning, 9(1/2), 9-29. Retrieved December 27, 2013, from EBSCO Online Database Education Research Complete. http://search.ebscohost.com/login.aspx?direct=true&db=ehh&AN=85423610&site=ehost-live

Beck, J., Stern, M. & Haugsjaa, E. (2004). Applications of AI in education. Crossroads: The ACM Student Magazine, 14. Retrieved October 31, 2007, from Association for Computing Machinery http://www1.acm.org/crossroads/xrds3-1/aied.html

Gong, Y., Beck, J. E., & Heffernan, N. T. (2011). How to construct more accurate student models: Comparing and optimizing knowledge tracing and performance factor analysis. International Journal of Artificial Intelligence In Education, 21(1/2), 27-46. Retrieved December 27, 2013, from EBSCO Online Database Education Research Complete. http://search.ebscohost.com/login.aspx?direct=true&db=ehh&AN=70210291&site=ehost-live

Johnson, D. (2005). Computer tutors get personal. Learning & Leading with Technology, 33 , 14-23. Retrieved November 4, 2007 from EBSCO Online Database Education Research Complete. http://search.ebscohost.com/login.aspx?direct=true&db=ehh&AN=18907237&site=ehost-live

Karlin, S. (2007). Futurist Liebowitz looks at tomorrow's schools today. American School Board Journal, 194 , 36. Retrieved October 30, 2007 from EBSCO Online Database Education Research Complete. http://search.ebscohost.com/login.aspx?direct=true&db=ehh&AN=26215122&site=ehost-live

Magliaro, S., Lockee, B., & Burton, J. (2005). Direct instruction revisited: a key model for instructional technology. Educational Technology Research & Development, 53 , 41-55. Retrieved November 4, 2007 from EBSCO Online Database Education Research Complete. http://search.ebscohost.com/login.aspx?direct=true&db=ehh&AN=19511443&site=ehost-live

Montalvo, G. (2006). Online design elements: Improving student success and minimizing instructor load. Mid-Western Educational Researcher, 19, 35-39. Retrieved November 4, 2007 from EBSCO Online Database Education Research Complete. http://search.ebscohost.com/login.aspx?direct=true&db=ehh&AN=19719837&site=ehost-live

Ong, J. & Ramachandran, S. (2000). Intelligent Tutoring Systems: The what and the how. Retrieved October 31, 2007, from the ASTD http://www.astd.org/LC/2000/0200_ong.htm

Sessink, O. Beeftink, H.,Tramper, J., & Hartog, R. (2007). Proteus: A lecturer-friendly adaptive tutoring system. Journal of Interactive Learning Research, 18, 533-554. Retrieved October 30, 2007 from EBSCO Online Database Education Research Complete. http://search.ebscohost.com/login.aspx?direct=true&db=ehh&AN=25977794&site=ehost-live

VanLehn, K. (2011). The relative effectiveness of human tutoring, intelligent tutoring systems, and other tutoring systems. Educational Psychologist, 46, 197-221. doi:10.1080/00461520.2011.611369. Retrieved December 27, 2013, from EBSCO Online Database Education Research Complete. http://search.ebscohost.com/login.aspx?direct=true&db=ehh&AN=66646074&site=ehost-live

Wijekumar, K. (2006). Implementing web-based intelligent tutoring systems in K-12 settings: a case study on approach and challenges. Journal of Educational Technology Systems, 35 , 193-208. Retrieved November 2, 2007 from EBSCO Online Database Education Research Complete. http://search.ebscohost.com/login.aspx?direct=true&db=ehh&AN=23893640&site=ehost-live

Suggested Reading

Ainsworth, S., & Fleming, P. (2006). Evaluating authoring tools for teachers as instructional designers. Computers in Human Behavior, 22, 131-148. Retrieved November 4, 2007 from EBSCO Online Database Education Research Complete. http://search.ebscohost.com/login.aspx?direct=true&db=ehh&AN=18730372&site=ehost-live

Milheim, W. (2006). Strategies for the design and delivery of blended learning courses. Educational Technology, 46 , 44-46. Retrieved November 2, 2007 from EBSCO Online Database Education Research Complete. http://search.ebscohost.com/login.aspx?direct=true&db=ehh&AN=23341699&site=ehost-live

Moore, D. (2006). The technology/inquiry typology: defining instructional technology. Journal of Interactive Learning Research, 17 , 401-406. Retrieved November 2, 2007 from EBSCO Online Database Education Research Complete. http://search.ebscohost.com/login.aspx?direct=true&db=ehh&AN=22079489&site=ehost-live

Scandura, J. M. (2012). Comments on Ansari & Sykes and Gogus, and suggestions for future research. Technology, Instruction, Cognition & Learning, 9(1/2), 51-56. Retrieved December 27, 2013, from EBSCO Online Database Education Research Complete. http://search.ebscohost.com/login.aspx?direct=true&db=ehh&AN=85423612&site=ehost-live

Troussas, C. C., & Virvou, M. M. (2013). Information theoretic clustering for an intelligent multilingual tutoring system. International Journal of Emerging Technologies In Learning, 8, 55-61. doi:10.3991/ijet.v8i6.3235. Retrieved December 27, 2013, from EBSCO Online Database Education Research Complete. http://search.ebscohost.com/login.aspx?direct=true&db=ehh&AN=92722277&site=ehost-live