This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.
This chapter will contain the review of literature of previous research that is considered significance in the development of Intelligent Tutoring System for primary school students.
2.1 Problem Domain
Problem domain is the area that needs to be examined to solve a problem. In this project, Intelligent Tutoring System is used in the domain of education. Education is the field for learning and teaching. It is the process where knowledge is transferred and received.
The purpose to have Intelligent Tutoring System for this project is because to give an output learning to the students especially primary school students. Nowadays, teachers or educators often face difficulties to handle their students. It is because one teacher needs to cater many classes and each class will have around 20 to 30 students. It is impossible to cater each student needs and preference.
Each student have their own learning style either they are good in listening, visualizing, or doing material at hand. Since the teacher is impossible to know each student learning style, therefore, there are needs for Intelligent Tutoring System that can provides a tutoring system that can determines the students interest so they will not having problems such as lack of understanding and misconceptions. Besides that, the benefit of this tutoring system towards the teacher is the teacher will easily monitor the student's performance and they will know the suitable approach to deal with the student styles of learning.
2.1.1 Learning Styles
Students learn in many ways such as by seeing and hearing, reflecting and acting, reasoning logically and intuitively, memorizing and also visualizing (Felder, n.d.). Teaching and learning are different between each person. It's all depends on the individuals itself. Everybody has learning style strengths in which different people will have different strengths (Dunn, 1990).
In 1986, Marie et al., as cited in (Farwell, 2000) provided an analysis in which about 20 to 30 percent of the school-aged students remembers what is heard, 40 percent easily recalled what they have seen or read and the rest were usually used both techniques which is they heard and visualize at the same time. They have their own way that will help them in learning.
There are several different theories concerning learning styles. Auditory, kinesthetic and visual are three types of fundamental learning styles (Graham, n.d.). Below are the descriptions for each learning styles as cited in Graham (n.d.) and Farwall (2002).
Children who are auditory learner usually prefer more on listening to explanation by reading them and sometimes they like to study by reciting information aloud. Furthermore, auditory learner may love to surround with music while studying or they may need a quiet space to study without diverted with any sounds. Auditory learners learn successfully when the way of giving information are being spoken and presented verbally.
"Show me and I will understand" is the keyword for visual learning style. It is a practice to reach new information by looking at something and visualize it. Usually, those people with this kind of learning style can grab information presented in chart or graph, but they may nurture impatiently listening to an explanation.
Most of the school's children excel through kinesthetic which means touching, feeling, and experiencing the material at hand. Learning activity such as science lab, field trip, skit and many other activities are the best technique for kinesthetic learner.
Most people use the combination of style to get best learning style for themselves. As for this project, there are two learning styles covered which is the visual and auditory learning styles. This learning styles can be classify via some set of personality questions in which it will determine the students area of learning styles.
Intelligent Tutoring System will usually come across to several techniques such as Case Based Reasoning technique, Agents technique, Neural techniques, Neuro Fuzzy techniques, Track Analysis and many more. Below are other potentials techniques that can be used for this Intelligent Tutoring System's project despite the Agents technique.
2.1.2 Examples of using Bayesian Networks for Learning Styles Detection
Bayesian Networks is one technique that detects student's learning styles in a web-based education system. In 2005, Garcia et al., had proposed this technique to ensure that all the students can learn even though they have different learning styles. Furthermore, Garcia et al. (2005), had also stated that intelligent agent can used those information to gives the students personalized assistance and deliver teaching components that suit best according to student's learning styles.
Table 2.1 shows the dimensions of the learning styles. Sensors like specifics data and testing; intuitive prefer ideology and theories. Sensors are enduring with detail but do not like complications; intuitive are uninterested by detail and love complications.
Table 2.1 Dimensions of Felder's learning styles
(Source: Gracia et al.,(2005))
A Bayesian Networks (BN) is a directed acyclic graph encodes the dependence relationships between a set of variables (Pardalos, n.d.). It allows us to discover new knowledge by combining expert domain knowledge with statistical data. In this BN, the nodes represent the different variables that determine a given learning style. The arcs represent the relationships among the learning style and the factor determining it. As shown in Figure 2.1, the model only has the three dimensions of Felder's framework, perception, processing and understanding.
Figure 2.1 Bayesian Network modelling student's learning styles.
(Source: Gracia et al.,(2005))
The following sentences describe in detail the different states the independent variables can take:
Forum: post messages; replies messages; reads messages; no participation.
Chat: participates; listens; no participation.
Mail: uses; does not use.
Information access: in fits and starts; sequential.
Reading material: concrete; abstract.
Exam Revision (considered in relation to the time assigned to the exam); less than 10%; between 10 and 20%; more than 20%.
Exam Delivery Time (considered in relation to the time assigned to the exam); less than 50%; between 50 and 75%; more than 75%
Exercises (in relation to the amount of exercises proposed): many (more than 75%); few (between 25 and 75%); none
Answer changes (in relation to the number of questions or items in the exam): many (more than 50%); few (between 20 and 50%); none.
Access to examples (in relation to the number of examples proposed): many (more than 75%); few (between 25% and 75%); none
Exam Results: high (more than 7 in a 1-10 scale); medium (between 4 and 7); low (below 4).
The probability functions associated with the independent nodes are gradually obtained by observing the student interaction with the system. 30 Computer Science Engineering students have been interviewed to determine the values experimentally using the ILS (Index of learning styles) questionnaire. Then, let the students used the education system and recorded their interactions with the system. The information was used to determine the parameter of the BN.
The Bayesian model is continuously updated as new information about the student's interaction with the system is obtained. The probability functions are adjusted to show the new observations or experiences. The probabilities reach equilibrium at certain point in the interaction. The probability values show a very small variation as new information is entered. The values obtained at this point represent the student's behavior.
This paper considered for each dimensions three values to make the results more comparable. For example, for the understanding dimension, it considered the values sequential, neutral and global. The percentage of coincidences is 100% for the understanding dimension, 80% for the perception dimension, and 80% for the processing dimension. All information from this paper is cited from Gracia et al, (2005).
In this project, Intelligent Tutoring System is used to classify students learning styles. It used numbers of rules as the main technique because it has the potential to give appropriate output which is the learning styles for the students. Below are the descriptions of all techniques that will be used in this project.
2.2.1 Intelligent Tutoring System
An early outline of Intelligent Tutoring System (ITS) requirements was delivered by Hartley and Sleeman in 1973 (Shute & Psotka, n.d). As stated by Shute and Psotka, Hartley and Sleeman argued that ITS must possess knowledge of the domain (expert model), knowledge of the learner (student model), and knowledge of teaching strategies (tutor). Furthermore, in order for ITS to have appropriate control strategies, it need to have captivating environment of learning, effectiveness of communication and to have flexible decisions. The ITS is a program in which student can communicated through a sequence of natural language questions and answers and the tutor could both ask and answer questions and keep track of ongoing dialogue structure (Corbett, Koedinger & Anderson, 1997).
The classic ITS architecture consist of four components which are a task environment, a domain knowledge module, a student model and pedagogical module.
Figure 2.2 ITS architecture
(Source: Corbett, Koedinger & Anderson, (1997))
As cited in Corbett, Koedinger & Anderson (1997), students engage in problem solving environments and these actions are evaluated with respect to the domain knowledge components. Student's knowledge state is maintained based on the evaluation model. Finally, the pedagogical module delivers instructional actions based on the evaluation of student's actions and on the student model.
Advantages of ITS as described by Yousoof, Sapiyan & Kamaludin (2002), ITS is a systems that can provide considerable flexibility in presentation of material and greater adaptability to respond to idiosyncratic students need. It also found to be highly effective in their purpose. It has been proved by research that the students who tend to learn using ITS actually could learn fast when compared to the students using traditional way of teaching.
Disadvantages of ITS as also cited in Yousoof, Sapiyan & Kamaludin (2002), risk issues affects the implementation of ITS, unsuccessful ITS can cause the barrier in implementation of ITS, replacement of human tutor will also be a barrier in implementation and wide spread of ITS will take place in another five years.
2.2.2 Rule Based Expert System
Expert system is a computer program that uses knowledge and inference procedure to solve problem that are difficult enough to require significant human expert to solve for their solution (Negnevitsky, 2002). It is also a computer program in which it is able to perform at the level of a human expert in a fine problem area. The most popular expert systems is a rule based expert system. It also called as production rules in which it contains IF-THEN statement.
Structure of Rule Based Expert System
A rule-based expert system has five components: the knowledge base, the database, the inference engine, the explanation facilities, and the user interface.
Figure 2.3 Basic Structure of Rule Based Expert System
(Source: Negnevitsky, (2002))
The knowledge base contains the domain knowledge useful for problem solving. In a rule-based expert system, the knowledge is represented as a set of rules. Each rule specifies a relation, recommendation, directive, strategy or heuristic and has the IF (condition) THEN (action) structure. When the condition part of a rule is satisfied, the rule is said to fire and the action part is executed.
The database includes a set of facts used to match aligned with the IF (condition) parts of rules stored in the knowledge base.
The inference engine brings out the reasoning whereby the expert system reaches a solution. It links the rules given in the knowledge base with the facts provided in the database.
The explanation facilities enable the user to ask the expert system how a particular conclusion is reached and why a specific fact is needed. An expert system must be able to explain its reasoning and justify its advice, analysis or conclusion.
The user interface is the means of communication between a user seeking a solution to the problem and an expert system.
The user is the one who will be used the system. User is also the one that will seek for solution.
Advantages of Rule Based Expert System
Natural knowledge representation.
An expert usually explains the problem solving procedure with such expressions as this: 'in such-and-such situation, I do so-and-so'. These expressions can be represented quite naturally as IF-THEN production rules.
ii. Uniform Structure.
Production rules have the uniform IF-THEN structure. Each rule is an independent piece of knowledge. The very syntax of production rules enables them to be self-documented.
iii. Separation of knowledge from its processing.
The structure of a rule-based expert system provides an effective separation of the knowledge base from the inference engine. This makes it possible to develop different applications using the same expert system shell. It also allows a graceful and easy expansion of the expert system. To make the system smarter, a knowledge engineer simply adds some rules to the knowledge base without intervening in the control structure.
Disadvantages of Rule Based Expert System
Opaque relations between rules
Although the individual production rules tend to be relatively simple and self-documented, their logical interactions within the large set of rules may be opaque. Rule-based systems make it difficult to observe how individual rules serve the overall strategy. This problem is related to the lack of hierarchical knowledge representation in rule based expert systems.
ii. Ineffective search strategy
The inference engine applies an exhaustive search through all the production rules during each cycle. Expert systems with a large set of rules (over 100 rules) can be slow, and thus large rule-based systems can be unsuitable for real-time applications.
Inability to learn
In general, rule-based expert systems do not have an ability to learn from the experience. Unlike a human expert, who knows when to 'break the rules', an expert system cannot automatically modify its knowledge base, or adjust existing rules or add new ones. The knowledge engineer is still responsible for revising and maintaining the system.
All information for Rule Based Expert System is cited from Negnevitsky (2002).
2.2.3 Intelligent Agent
An agent is anything that can be viewed as perceiving its environment through sensors and acting upon that environment through effectors (Rusell & Norvig, 1995). Currently, agents are the point of interest on the part of many areas of Computer Science and Artificial Intelligence.
According to Jennings & Wooldridge (n.d.), an intelligent agent is a computer program that is able to perform immediate response in order to meet its design objectives. Flexible here means that the systems must be responsive in which agents should distinguish their environments and react in a well-timed to changes that occur in it. Agents should also be proactive whereby they should be able to exhibit opportunities, goal-directed behavior, and take the initiative where appropriate. Finally, agents should be social in which agents should be interrelate when they comfortable with other Artificial Agents and humans in order to complete their own problem solving and to help others with their activities.
Advantages of using Intelligent Agent are because agents represent a powerful tool for making system more flexible. Agents should behave like an 'expert assistant' with respect to some application, knowledgeable about both the application and the user, and capable of acting with user in order to achieve the user's goals. Agents are also good in improving the efficiency of Software Development.
The limitations or the disadvantages of using agent as discussed by Jennings & Wooldridge (n.d.) are:-
No overall system controller
An agent-based solution may not be suitable for domains in which global constraints have to be maintained, domains where a real-time response must be guaranteed, or in domains in which deadlocks or live locks must be avoided.
No global perspective
Agents may make globally sub-optimal decisions since in almost any realistic agent system; complete global knowledge is not a possibility. An agent's action are by definition determined by that agent's local state.
Trust and delegation
Users have to gain confidence in the agents that work on their behalf, and this process can take some time. During this period of time, the agent must strike a balance between continually seeking guidance (and needlessly distracting the user) and never seeking guidance (an exceeding its authority). An agent must know its limitations.
2.2.4 Multiagent System
As stated by Capuano et al. (n.d.), multiagent system (MAS) can be defined as loosely-coupled networks of communicating and cooperating agents functioning together to solve problems that are ahead of their individual capabilities. In order to obtain coherent system behavior, individual agents in a multiagent system are not only able to share knowledge about problems and solutions, but also to reason about the processes of coordination among other agents (Capuano et al., n.d.).
The idea of multiagent system is that an agent is a computer program that has capability to perform autonomous action on behalf of its owner or user. In addition, agent can figure out for itself what it needs to do in order to satisfy its design objectives. A multiagent system is one that consists of number of agents, which interact with another, typically by exchanging messages through some computer infrastructure (Wooldridge, 2002). In order to successfully interact, these agents will thus require the ability to cooperate, coordinate and negotiate with each other.
2.2.5 Distributed Case Based Reasoning
Case Based Reasoning (CBR) is another technique that is widely used in Intelligent Tutoring System and in the field of education indeed. As proposed by Rishi et al. (2007), they combine both technique which are CBR and agent technique to provide student modeling for online learning in a distributed environment with the help of agents.
In this paper, it focused more on Case Based Distributed Student Modeling (agent based) ITS architecture to support student-centred, self-paced, and highly interactive learning. The first step is to build the effective learning environment which is the CBR where the system maintains a complete and full set of cases (scenarios) of student's learning pattern and employs an efficient and flexible case retrieval system.
The system as cited in Rishi et al. (2007) must used the student's learning profile such as learning style and background knowledge in selecting, organizing and presenting the learning material to support case based learning. As Rishi et al. (2007) cited from Yi Shang et al. (2001) and Kumar (2005), Distributed CBR based student modeling enables adaptive delivery of educational contents and facilitates automatic evaluation of learning outcomes.
This system consists of three agents with different expertise. The first agent which is personal agent will focus on student profiler which include knowledge background, learning style, interests, course enrolled etc. The other two agents communicated with each other through different communication channel which situated in distributed environments are teaching agent and course agent. Figure 2.4 show the communication model among agents.
Figure 2.4 Communication model among agents
(Source: Rishi et al. (2007))
Furthermore, the following activities as shown in figure 2.5 take place during the student modeling when the student interacts with the system as such, selection of topic by the student and get to know student's background by presenting problems to the student, analyzing the student's response by the system, selection of case by the system based on response, adaptation of the case by the system, achieving the knowledge component of the student model through case retrieval, generation of teaching strategy by the system and presenting the next problem to the student.
Figure 2.5 Process of Student Modeling
(Source: Rishi et al. (2007))
Finally, this system is fully distributed in which it does not bounded with any network topology, it reduces the need of large storage spaces at the user's site to store all the cases and redundancy is maintained for fault tolerance. The whole system is managed in the distributed environment with just three agents which are Personal Agent, Teaching Agent and Course Agent.
2.2.6 Track Analysis
Learning indicators is the way of exploiting and analyzing the tracks and providing knowledge on the activities (Bousbia et al., n.d.). This will help teachers to perceive and interpret the learner's activities in e-learning situations. As in figure 2.5 this paper by Bousbia et al. (n.d.) considers three steps in the analysis.
The first one is indicator's choice. The first step is basically to guide the collection process. It helps the teacher to choose high level indicators in which the teacher intends to seek from the indicators base. It will then ask the teacher to provide additional data required for their calculations. The next step is observation. In this stage, the system identified the necessary tracks extracted through a collection tool which installed on the learner side. This tool has specific history such as visited pages URLs, time and actions. Finally, the analysis and interpretation step. This is the most important step in which it divided into three main stages which are browsing path rebuilding, indicators' calculation and learning style deduction.
Figure 2.6 Learning Style Deduction Steps
(Source: Bousbia et al. (n.d.))
There will be three layers remain which are educational preference layer, learning process layer and cognitive abilities layer. The first layer includes attributes related to the preferred learning time, environment preference, information representations and encoding methods. The second layer includes learning strategy, comprehension and progression approach. For the last layer, it includes motivation and concentration capacity.
The learning style can be determined by calculating the value of each layer's attribute. By using the necessary high level indicators, the value is deduced. Furthermore, to connect the indicators to the learning styles, Bousbia et al. (n.d.) classify them according to model layers. The possible values of each layer's attribute are chosen from the existing learning style models, by making their definitions closer.
2.3 Related Works
Related works are works from other researchers which have related to this project or perhaps the same technique used but in different field or domain. Intelligent Tutoring System and some other techniques is the main focused in this research to compare and differentiate domain and techniques with other projects.
2.3.1 Intelligent Agent in E-commerce
Ecommerce or e-commerce is the ability and skill of selling products or services over the Internet (Ward, 2010). As discussed by Pivk & Gams (n.d.) in their article on Intelligent Agent in E-commerce, the article discussed on assessment of agent technologies which involved in buying and selling. Several agent-mediated electronic commerce systems are analyzed in the perspective of a general model of the buying process.
E-commerce involves business-to-business (B2B), business-to-customer (B2C) and customer-to-customer (C2C) transactions. It encounters a wide range of issues such as security, trust, electronic product, catalogues and many more. Intelligent agent can be used or applied to any of these.
Pivk & Gams (n.d.) had given examples on the usage of agent in ecommerce such as Tete-a-Tete ([email protected]). For examples in Figure 2.7, a shopping agent may receive proposals from multiple sales agents. Each proposal defines a complete product offering including a product configuration, price and the merchant's value-added services. The shopping agent evaluates and order these proposals based on how well they satisfy its owner's preferences. If the shopper is not satisfied, he can critique them along one or more dimensions. User shopping agent broadcasts this preference changes to the sales agents in which, in turn, use them to counter-propose better product offering.
Figure 2.7 Consumer-owned shopping agents integrative negotiate with multiple merchant-owned sales agents.
(Source: Pivk & Gams (n.d.))
2.3.2 Intelligent Agent Based Graphic User Interface (GUI) for e-Physician
This paper is proposed by Jung, Thapa & Wang (2007). It is all about the approach of using ontology based intelligent interface agent that will assist the physician to acquire online access interface to patient's chart, fast rescheduling such as emergency case, easy access to laboratory results and reducing overall cost because of optimum utilization of time.
In this paper, medical homecare system framework is designed in real-time environment. There are four types of agents that are used in this system which are Interface Agent, Admin Agent, Laboratory Agent, Diagnosis Agent and Schedule Agent as in Figure 2.8.
Figure 2.8 Conceptual Framework of Intelligent Agent
Based user interface
(Source: Jung, Thapa & Wang (2007))
As cited from Jung, Thapa & Wang (2007), interface agent is the agent that will interact with the user and will work as an information filtering agent and select the most critical case per priority. On the other hand, laboratory agent will be able to provide the detail examination report from the laboratory database. Furthermore, diagnosis agent will help the interface agent to suggest proper diagnosis by using decision making rules. In addition, administrative agent will provide pre-historic diagnosis trend of the patient and finally schedule agent will help in preparing patient's chart, scheduling, and fat rescheduling of the plan on basis priority. Those agents will help in the development of the system and will give user's final control for optimization of their best GUI.
2.3.3 Intelligent Agent in Computer Games
Games are the virtual worlds that are more traceable than real world. It is also something that can be controlled, formal, and measurable, provide realistic and significant challenge (Mikkulainen, n.d.). Intelligent agents can be deployed in games today.
As cited in Lent et al. (n.d.), they discussed on the use of intelligent agent in the games called Soar. It make the growth of intelligent agents for games easier by giving common inference engine and reusable knowledge base that can be simply applied in many different games. Soar allows easy decomposition of the agent's action through hierarchy of operation. It used Quake II and Descent 3 agents in which both have the functionality in the games such as flying in a spaceship without gravity, attack, explore and many more.
Furthermore, Soar constantly cycles through perceive in which it accept sensor information from the game, think (select and execute relevant knowledge) and Act (Execute internal and external actions). Interface is another important part in developing games since the interface extracts the necessary information from the game and encodes it into the format required in Soar.
2.3.4 Neural Network-based Fuzzy Modeling of the Student in ITS
This paper is using empirical approach that use the neuro-fuzzy synergism to evaluate the students in the context of an ITS is presented. Stathacopoulou, Magoulas & Grigoriadou (n.d.) stated that fuzzy logic techniques is widely used in ITS since it have the ability to handle imprecise information such as student's actions and to provide human descriptions of knowledge and of student's cognitive abilities.
In this paper, fuzzy logic is used to provide human-like approximate diagnosis of student's knowledge and cognitive abilities and Neural Network is used to trained human teacher's decisions regarding student's characteristics and fixed weight Neural Network are used to evaluate and aggregate membership function.
The neuro-fuzzy model has been tested in physics domain to evaluate student's characteristics for deciding about the appropriate teaching strategy. Experiments have been performed by Stathacopoulou, Magoulas & Grigoriadou (n.d.) using a population of 300 simulated student cases with the decisions of 5 teachers. The overall average classification success has been 95 %. As conclusion, evaluation of students depends on designer's ability to analyze the cognitive domain suitably, define fuzzy sets and relate the student response with suitable knowledge and cognitive characteristics.
2.3.5 FlexiTrainer: A Visual Authoring Framework for Case-Based Intelligent Tutoring System
FlexiTrainer is an authoring framework that allowed the fast swift design of pedagogically rich and performance-oriented learning environments with tradition content and tutoring strategies (Ramachandran, Remolina & Fu (n.d.)). This authoring tool specifies a dynamic behavior of tutoring agents that interact to deliver instruction. FlexiTrainer has been used to develop an ITS for training helicopter pilots in flying skills.
FlexiTrainer consists of two components which are the authoring tools and the routine engine. Core component for FlexiTrainer are Task-Skill-Principle Editor, Exercise Editor, Student Model Editor, and Tutor Behavior Editor. Each of these editors has their own specific functionality. An instructional agent is used to carry out teaching-elated to achieve instructional goals. It used Bayesian inference to incorporate student modeling strategies.
2.3.6 Intelligent Tutoring System using Hybrid Expert System with Speech Model in Neural Networks
This paper used supervised learning neural networks to successful rate. Besides being more information delivery systems, this system help students to actively construct knowledge. This paper by Venkatesh, Naganathan & Maheswari (2010) enable teaching system to be developed in various fields and subjects.
Neural Model in this system is used for Question Answering System. As shown in Figure 2.9, input layer contains the questions on the wanted subjects. Both possible questions and answers are stored in the desired output.
Figure 2.9: Neural Network Architecture
(Source: Venkatesh, Naganathan & Maheswari, (2010))
On the other hand, speech model consists of language extraction (includes classes such as noun, verb, operator, pronoun and many more), speech act classifier (tutor uses strings of words and punctuation to classify each contribution of the learner into speech act units), file management (used as marker for the lecturer's mode answer which communicated with ITS module) and modes (select person to be communicated with the ITS either student, lecturer or admin). This system not only reduce development times but also appreciably simplifies the technical knowledge required of personnel involved in the generation of an auto-regulated intelligent tutoring dialogue system (Venkatesh, Naganathan & Maheswari (2010)).
There are many ways in developing Intelligent Tutoring System as mentioned above. Each technique used has its own strengths and weaknesses. In this project, Rule-based is used since it gives more impact and significance to the prototype.
The next chapter will show the research framework on the methodology for developing this prototype.