Course:COGS200/2017W1/ProjectGroup22
Full Proposal
Introduction
Our group 22 proposes a project based on ITS (Intelligent Tutoring Systems) and its role on educational development. Before that we have some various ideas for this project such as Effects of Yoga on Brain Plasticity and Aging, however, we have decided to go with ITS because it is more relevant to our course as it can be useful for many students in our class. It is useful to get some additional knowledge on ITS for the reason that it covers the part of AI which is one of the most important topics in COGS200 and in the field of studies such as Computer Science and Philosophy. It also plays its role in the educational system as ITS can be the essential part of student's life all over the world in the future. The purpose of an ITS is to provide independent feedback, interactive help and do both assessment and plan recognition (what a student already knows and what a student is trying to achieve in learning). By integrating educational psychology and artificial intelligence, we are able to gain a foresight of the extent to which intelligent systems contribute to the future of learning. The linguistics aspect enables us to map how a specific demographic picks up on foreign language if taught by an intelligent system, as compared to existing studies involving different learning methods. We have found plenty of relevant background research articles to support our hypothesis which are mostly based on Bayesian Networks which help to manage diverse difficulties and struggles with student-modelling system in ITS.
Methods
This section will describe your approach to solve the problem or address the question laid out in the introduction. How do you propose to solve the problem/hypothesis? This could include the specific experimental or linguistic approach you will take to solve a psychological, computational, philosophical or linguistic issue. If you are designing an artificial system, you will discuss the modules and/or algorithms that you will employ to solve your problem.
Our group proposes a project in which we conduct an experiment involving two groups of students in the same age range. The first group, the control group, is to be taught in a given subject (i.e. language) by a typical teacher, whereas the experimental group is to be taught in the same subject, but by an intelligent system. Afterwards, both groups are to be administered a test, the results of which will be compared to determine the efficiency of an intelligent system in teaching. By integrating the fields of computer science, educational psychology and a bit of Linguistics in this study, we are granted a foresight into the extent to which artificial intelligence contributes to the future of learning. Several intelligent tutoring systems have already been implemented today, the research behind which we will utilize to support our project. This study will consist of monolingual English-speaking students, spanning a similar age range across both groups. To account for third-party variables, this study will be exclusive to students who do not suffer from any form of test anxiety. Over the span of 12 weeks, the students will be taught a language course that they have no prior knowledge of (ie. French or Spanish) by their respective tutors. Once the 12 weeks are up, both groups will be administered a test to assess their understanding of the material they have just been taught. The test will include both an oral portion as well as a written portion. Afterwards, we will calculate the results and compare to see if there are any substantial differences between the two groups.
We checked different web based AI tutors online and finally decided to go with eTeacher AI tutor. The different web based AI tutors we checked are listed under reference.
In this paper we present eTeacher, an intelligent agent that provides personalized assistance to e-learning students. eTeacher observes a student’s behavior while he/she is taking online courses and automatically builds the student’s profile. This profile comprises the student’s learning style and information about the student’s performance, such as exercises done, topics studied, exam results. In our approach, a student’s learning style is automatically detected from the student’s actions in an e-learning system using Bayesian networks. Then, eTeacher uses the information contained in the student profile to proactively assist the student by suggesting him/her personalized courses of action that will help him/her during the learning process.
One of the most desired characteristics of e-learning systems is that of being personalized, since they have to be used by a wide variety of students with different skills. Different students learn in different ways. Some of them process information reflectively while others actively. Some students prefer abstract material while others prefer concrete examples. Some study steadily while others in fits and starts. Thus, to be effective, elearning systems should consider each student’s learning preferences and skills. Intelligent agents are computer programs that learn users’ interests, preferences, and habits and give them proactive, personalized assistance with a computer application. In this work, we present eTeacher, an agent that provides personalized recommendations to students depending on their profile and on their performance with a certain Web-based course. To achieve its goal, eTeacher has to build first a student profile. In eTeacher, the student profile is given mainly by the student’s learning style. A learning style model classifies students according to where they fit on a number of scales belonging to the ways in which they receive and process information. The use of learning styles for experimental research in Web-based educational systems has demonstrated that providing material according to students’ learning styles can enhance students’ learning and that these styles are linked to quantitative differences in both navigation behavior and learning performance.
To provide personalized assistance to students, eTeacher must build a student profile. To achieve this goal, eTeacher unobtrusively observes the student’s behavior while he/she is taking a course via an e-learning system. The agent records the student’s actions and then it uses these data and the data logged by the system to build the student profile. As said before, the student profile comprises the student’s learning style and the student’s performance with a given course, such as the number and type of exercises done, the topics studied, and the results in exams. The student’s learning style is determined by a Bayesian network. Bayesian networks enable eTeacher to model quantitative and qualitative information about a student’s behavior with the e-learning system. The agent can infer the student’s learning style using the Bayesian network given different student behaviors observed. For example, if a student participates in chat rooms and forums, eTeacher can infer that the student processes information actively and not reflectively. Then, when eTeacher detects situations in which the student might need assistance or guidance, it provides him/her help according to the student’s learning style, his/her stage in the course, and his/her performance. After eTeacher’s recommendations, the student can provide feedback to the agent’s assistance. This feedback can be explicit, through a user interface provided for this purpose, or implicit, if the agent observes the student’s actions after assisting him/her. In turn, the feedback provided can be positive, when the student accepts the agent’s suggestions, or negative, if the student rejects the agent’s assistance. eTeacher uses this feedback to adjust the information it has about the user and act accordingly in the future. Fig. 1 shows an overview of the agent’s functionality.
Bayesian networks (BN) enable eTeacher to model both qualitative and quantitative information about students’ learning styles. A BN is a compact, expressive representation of uncertain relationships among variables of interest in a domain. A BN is a directed acyclic graph that represents a probability distribution, where nodes represent variables and arcs represent probabilistic correlation or dependency between variables. The strengths of the dependencies are given by probability values. For each node, a probability table specifies the probability of each possible state of the node given each possible combination of states of its parents. These tables are known as conditional probability tables (CPT). Tables for root nodes (or independent nodes) just contain unconditional probabilities. In our agent, random variables represent the different dimensions of Felder’s learning styles and the behaviors that determine each of these dimensions. These behaviors are extracted from the interactions between the student and the e-learning system. Table 2 shows the behaviors that help the agent determine each of the dimensions. This information is obtained by analyzing the data recorded in a student’s log file. With respect to the Perception dimension we can say that, according to Felder, a student who does not revise his/her exercises or exams is likely to be intuitive. On the other hand, a student who carefully checks the exams or exercises is generally sensory. A student who reads or accesses various examples of a given topic is more sensory than one who reads just one or two. As regards the type of reading material the student prefers a sensory learner prefers concrete (application oriented) material, while an intuitive learner usually likes abstract or theoretical texts. To detect whether the student prefers to work things out alone (reflectively) or in groups (actively), we analyze his/her participation in forums, chats, and mail systems. As regards forums, we analyze whether the student begins a discussion, replies to a message, or just reads the messages posted by other students. The frequency of this participation is also important. The participation in chat and mails can give us some information, but it is not as relevant as the one we can obtain with a forum access log. To enable students to work collaboratively, most Web-based educational systems provide a collaborative facility. Given a certain problem solving activity, students use this tool to propose solutions to the problem, to make a counter-proposal to a proposed solution and to read the solutions available thus far. They can also send messages to other members of the group or read messages posted by others. In addition, the system logs the participation of each student in the group activity. Finally, to determine how students understand, we analyze access patterns to information, which are recorded in students’ log files. If the student jumps through the course contents we can say that he/she does not learn sequentially but in fits and starts. The results the student gets in the exams while he/she is jumping over the contents give us an indication of his/her understanding style. If the student gets a high mark in a topic despite having not read a previous topic, we can conclude that the student does not learn sequentially.
The dependencies between learning styles and behaviors are encoded in the Bayesian model through the arcs that go from the nodes representing student behaviors to the nodes representing learning style dimensions. Fig. 2 shows the Bayesian model used by our agent. The values the different variables can take are summarized below:
Forum: posts messages; replies messages; reads messages; no participation.
Chat: participates; listens; no participation.
Mail: uses; does not use.
Tasks: makes proposal for group task; makes counterproposal; reads proposal.
Messages: sends message; reads message (within group task).
Participation: participates; no participation.
Information access: in fits and starts; continuous.
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).
Fig. 2.
Initially, probability values for independent nodes are assigned equal values. Then, the values are updated as the system gathers information about the student’s behavior. The probabilities attached to the independent nodes are adjusted to represent the new observations or experiences (Olesen, Lauritzen, & Jensen, 1992). Consequently, the Bayesian model is continuously updated as new information about the student’s interaction with the system is obtained. On the other hand, the probability values contained in the different CPT were obtained via a combination of expert knowledge and experimental
results.
Once we have the Bayesian model, the goal of eTeacher is inferring the values of the nodes corresponding to the dimensions of a learning style given evidences of the student’s behavior with the system. Thus, we obtain the probability values of the learning style node given the values of the related independent nodes. The learning style of the student is the one having the greatest probability value.
Discussion
Why did you choose the method/approach that you did? What are the predictions of results if employing experimental approach? How will you evaluate the performance of your system if building software/hardware?
We have described eTeacher, an intelligent agent that assists e-learning students depending on their learning styles and on their performance with a Web-based course. eTeacher uses Bayesian networks to build the student profile. The agent has been successfully evaluated with real students and the results obtained are promising. In the future, we will further study students’ log files to obtain more information about eTeacher’s performance. As a future work, we are planning to incorporate the input dimension in the student’s model. Currently, eTeacher cannot distinguish between visual and verbal learners, and hence, it cannot provide them assistance accordingly. New suggestions and messages can be added to enhance eTeacher’s functionality. In addition, thus far, eTeacher recommends courses of actions according to the student’s learning style, favoring the advantages of each style. We are working now towards a different research direction, that is, the agent can suggest actions that tend to complement the learning styles. For example, if a student is intuitive we know that he/she does not like to revise his/her exam and might make mistakes, that the agent recommends him/her to revise the exam before handing it out. Then, we will be able to compare the two approaches: assisting students favoring their learning styles vs. complementing their learning styles.
Conclusion
What insights did you draw from this project? What did you learn from this experience?
Bibliography
Anderson, John R., et al. "Cognitive tutors: Lessons learned." The journal of the learning sciences 4.2 (1995): 167-207
Conati, C., Gertner, A., Vanlehn, K. (2002). Using Bayesian Networks to Manage Uncertainty in Student Modeling. ProQuest , 12
Gardner, Robert C. "Learning another language: A true social psychological experiment." Journal of language and Social Psychology 2.2-3-4 (1983): 219-239
Gass, Susan M. "Second language acquisition and linguistic theory: The role of language transfer." Linguistic theory in second language acquisition. Springer Netherlands, 1988. 384-403
Naser, S. "Evaluating the effectiveness of the CPP-Tutor an intelligent tutoring system for students learning to program in C++." Journal of Applied Sciences Research 5.1 (2009): 109-114
Ma, Wenting, et al. "Intelligent tutoring systems and learning outcomes: A meta-analysis." (2014): 901
http://www.eteachergroup.com/schools/
http://www.highwaytoenglish.com/en-us/how-highway-to-english-works/
https://pdfs.semanticscholar.org/f98d/91492335e74621837c01c860cbc801a2acbb.pdf
https://cseweb.ucsd.edu/~zzhai/blog/intelligent-tutoring-system-overview.html