Use of Web Technology for Interactive Remote Instruction
Abstract
Advancements in web technology are changing learning paradigms. In this paper we relate various learning paradigms to existing technology and describe two approaches in IRI to use web technology for synchronous sessions: web steering and control and automatic content synthesis using the web. IRI (for Interactive Remote Instruction) is a computer-based system to support distance education developed at Old Dominion University; it is being used to teach courses between sites up to 200 miles apart.
Higher education is undergoing structural changes in terms of composition of the student populations, learning paradigms used, and curricula. As distance learning is becoming an integral part of secondary institutions, their student body composition is expanding to include non-traditional students. Secondly, instructional methods in academia are shifting from a teacher-centered paradigm to a student-centered paradigm. In this new paradigm, the student becomes an active participant in the class, and peer collaboration becomes an important component in the learning process. The advances in computer networking and digital media technology, together with the growth of the Internet, are making virtual classrooms and web technology an effective framework for supporting active learning. In a virtual classroom the student interacts with other participants in learning activities using a desktop computer. In addition, this same computer can be used to support the active learning paradigm through its modeling capabilities. In this context we wish to examine three particular aspects:
Section 2 of this paper relates new learning paradigms and the technologies that support them. This section will also summarize IRI’s capabilities (Maly et al. 1996, 1997). Section 3 will describe IRI features supporting automatic content generation, and steering and monitoring of synchronous sessions. Section 4 provides a summary of IRI’s current status and future planned work.
The pace of learning paradigm shifts has accelerated over the last decade, particularly in the arena of higher education, due to the exponential evolution of communication and computer technologies. Restructuring, distance learning, and virtual classrooms are but a few concepts universities cannot ignore lest they become obsolete. This section describes how learning paradigms shift as a result of technological advances. First we shall summarize the relevant technologies, followed by a characterization of a learning paradigm, and then describe how various application level tools support a learning environment.
The technologies relevant to this subject are computers and their operating systems, application software, and network transport protocols. Today's computers are equipped with CPUs having processing speeds of at least 200MHz, and have the capability to capture and replay both video and audio. The two major operating systems are NT (including Windows95) and the Unix platforms. Together, these two probably cover 95% of all computers sold today. In terms of networks, there is a clear separation between local (LAN) and wide area networks (WAN). At this point, ATM seems to be preeminent for WANs and 10-100Mb/s Ethernets for LANs. It is doubtful that ATM will become an end-to-end solution because 1Gb/s Ethernet is the better (less expensive) candidate for LANs. The Internet, as it exists today, is based on the TCP/IP protocol suite, with multicast IP (and in the near future IPv6) becoming a major factor in applications which support reaching many users at a time.
We characterize a learning paradigm in terms of the following dimensions:
We discuss three higher education learning paradigms in terms of this characterization: a traditional class, a satellite-television based distance learning class, and a web based class.
Traditional class
The student enrollment of a traditional classroom ranges from five to 500, depending on the level of the course. The degree of symmetry is inversely proportional to class size. Hence, a graduate class of five students has a higher level of symmetry than a general introductory course with 500 students. During lecture time, synchrony and co-location are absolute, whereas asynchronous learning happens alone or in small, co-located groups (possibly involving a teacher or her assistants). Perception and interactivity are high due to the direct contact between the teacher and students. Tool usage during lectures is infrequent, perhaps involving demonstrations, and practically no sharing of tools is possible. In a laboratory session tool usage and sharing are more frequent, however, class size must be reduced to 15 to 20 participants. Cost is mostly determined by staff salaries and to a lesser degree by facilities. The cost for a laboratory session is high and is commonly found in engineering, health, and some science fields. The time of a course is mostly determined by accreditation agencies and term lengths, neither of which are in the student’s control, resulting in a low degree of time.
Distance learning class
A satellite-television based class (Fox 1995) consists of multiple sites, comprised of an origin site and multiple receiving sites. Site enrollment typically ranges from one to twenty, while class enrollment ranges from 10 to 100. Symmetry is very low because video is only sent from the originating site to the receiving sites and only audio channels are available for feedback from students. There is higher degree of time differential possible because TV broadcast can be recorded and played back at a time convenient for participants. Clearly, the student pays the price of not being able to interact through the audio channel. The degree of perception is low for any textual and many types of graphical material because of the limited display quality of TV (which will change somewhat with the introduction of high-definition TV). Interactivity is acceptable if low-orbit satellites are used. For all practical purposes, distance is unlimited. Tool usage and control by the participants co-located with the instructor are equivalent to that in the traditional lecture class. Tool use in laboratories is usually infeasible for remote sites unless they have the same facilities. Universities offering distance learning courses price them the same or lower than their on campus equivalents. Distance learning offers non-traditional students the possibility of earning a university degree, or just credit for specific courses. In this model of learning it is also possible to change the time it takes to complete a course. If mechanisms exist to handle assignments, projects, and exams, the videotapes of a course can be studied at a rate different from a traditional course.
Web-based class
Another learning model gaining widespread acceptance, either in its pure form or as a complement to other models, is the web course (Fox 1995). Web courses are related to the self-learning courses developed with the PLATO system in the late 1970s. In PLATO all the course material was stored on a computer and a program guided the student through the material at a self-regulated pace. PLATO never gained widespread acceptance due to it’s cost and limitations in the technology used. Today, with web technology, these problems have been largely solved. In web courses, the material is presented in multimedia hyperlinked form to the student; programs guide the student through the material based on an interactive assessment of the student's capability. If done properly this paradigm scores very high on the interactivity scale and can be taken totally asynchronous (i.e., whenever a particular student has time she can take a lesson). This model is not symmetric because it is only one student interacting with a program, unless the web course is enhanced by some technology that combines it with group interaction. The degree of tool use. Perception scores very high with today’s display technology of multimedia and colocation is not necessary. Anyone with access to the Internet can access the material assuming the student has been given access by registering. Cost for developing a good course can be prohibitive; many colleges have found that it costs on the order of $100,000 to develop a truly effective course. On the other hand, since this paradigm scales very well these costs can be amortized over many students. Another advantage of these courses is the flexibility it allows in setting time periods for achieving a learning objective.
The ideal paradigm we propose, and believe current technology can support, is one which scales well, is symmetric, allows for both asynchronous and synchronous modes, has high perception quality, high interactivity with a delay smaller than 50ms, allows for separation of students in space, allows for all computer learning tools to be used and used in a shared mode, is cost effective and can be taken over variable periods of time for a fixed set of learning objectives. The colocation and interactivity scales are not independent because the 50ms bound on the delay implies that participants cannot be farther apart than 10,000km.
We distinguish two types of tools: learning tools that help students learn a concept, skill or knowledge, and technology tools that enable such learning. In the first category we have seen, through the pervasive availability of computers, steady replacement of physical learning tools by computer tools. More and more, physical tools are not directly manipulated but rather are controlled and steered through a computer. For example, the electron microscope is actually controlled by a user interacting with a graphical user interface on a computer with the results displayed on a computer screen. Similarly, many physical laboratory experiments are being replaced by computer models that allow students to specify parameters for an experiment in which a program provides visual representations of results. Literally, tens of thousands of such programs have been written by educators for any conceivable disciplines.
The tools we describe in this section are the technology tools and environments that enable new learning paradigms. Technology tools support or enable a particular feature in a learning experience; for example, a video conferencing tool supports group interaction. An environment is a set of integrated tools that provide support for most of the learning experience of a student.
Collaboration Tools
Such tools enable, for a specific operating system, any learning program to be shared among a group of participants. Collaboration tools provide a means for any user to take control of a tool and operate it. One Unix example is XTV (Abdel-Wahab). It allows for the sharing of any X program. An NT example is NetMeeting (Microsoft). It enables a group to share any win32 program. Since these tools use TCP/IP, they do not scale well and cannot handle large groups. No tool currently exists which allows the sharing of arbitrary learning tools across platforms. However, there are a number of tools being developed that will allow users to share specific tools across all platforms. For instance, WEB-4M (JDH Technologies) is a JAVA based environment that provides specific tools, such as a whiteboard and presentation tool, which can be used on any platform. Again, this tool is based on TCP/IP and thus does not scale well. All of these tools suffer on both perception and interactivity scales when used over the Internet. The Internet, as it now exists, is unpredictably congested. Participants in a session may experience different levels of bandwidth available to them and different delays to interactions, depending on the time and location of their Internet connection. These differences can be of several orders of magnitude, making synchronization of group activities sometimes impossible. Even for one user, it is impossible to have a predictable synchronization of the multi-media streams (video, audio, data traffic) a participant receives.
Videoconferencing Tools
A plethora of tools exist, both over IP networks and ISDN based telephone networks, that support videoconferencing. Vic and vat (Network Research Group), a suite developed by Van Jacobson, was originally developed for large audiences as an asymmetric tool (i.e., mostly for broadcasting to thousands with very few sources). The tool is based on the Mbone (IP multicast) protocol and also includes a whiteboard. It is written for all platforms, although the NT version has some problems with the video. It is most commonly used over the Internet. Despite the Internets congestion (which gives the tool low interactivity and great variation in inter-stream delays), vic and vats price (free) makes it an attractive tool. SmartStation (VTEL) is a proprietary tool that works over the telephone network with switched ISDN lines. It does not scale very well because a central switch has to be able to handle all incoming ISDN lines. Being based on a fully connected switched infrastructure, it is symmetric; at any time anyone can become the focus of attention. Such a system can be expensive to operate. Typically at least three ISDN lines (192K) are necessary to provide each participant an acceptable level of perception. The product is usually bundled with a whiteboard for sharing graphical and textual information. None of the video conferencing tools support tool sharing.
Web Tools
These tools are used to support standalone web courses or as supplements to other learning paradigms. Web Browsers, such as Netscape and Internet Explorer, provide the simplest web tools. These include editors for creating multimedia web pages and groupware such as e-mail, chatrooms, newsgroups, and document managers. At the value-added level, Domino, a secure web server, provides a framework for writing applications for supporting activities within a group. It provides an ideal environment for handling all the administrative aspects of running a class, including access control, database access, and document management. Authoring tools have been developed to help a teacher, without the aid of specialists, transform the educational content of a course into an interactive web course. Specialized tools exist to support the assessment part of the learning process. For instance, QuestWriter (Oregon State University) is an interactive webtool which allows the teacher to create a quiz, give it in class (or have students take it at their own time), and have the tool grade the multiple-choice and fill-in-the-blank components automatically. On another note, JAVA is not only being used to allow students to interact with web pages, but it also providing students with learning tools. For example, the teacher might place a JAVA-written learning tool on the web that relates the force on an object with a trajectory. The student can execute the program with various parameters and observe the behavior of the object. Sharing of information and tools in a synchronous fashion is still in its infancy and is almost exclusively based on point-to-point TCP/IP connections, which have scaling limitations. The most common shared tools are whiteboards and specialized presentation tools. Co-Browsing is a coordinated browsing system that allows a group of participants to visit the same page on the web together. Anyone in the group can click on a link and all participants will see the result (Davis).
Cross-Platform Tools
One of the most vexing conundrums is the existence of two dominating operating platforms: NT and Unix. The identification of Unix with academia and NT with home PCs is steadily eroding. Yet, many learning tools only exist in one of the two worlds, hence the need for the ability to run a tool written for one platform on another platform. The best known tool for running X programs on a PC is Exceed, which basically creates an X server under NT and allows any X program running on a Unix box to be displayed and manipulated on the PC running NT. Citrix provides the opposite: it runs on an NT server and any program running under NT can be displayed and manipulated by a citrix client running on a Unix box. All of these tools, however, work on a one-to-one basis. That is, it is impossible for a group of users on Unix boxes to see a tool running on a PC.
Environments
A successful environment will require the end user to perform few operations to operate the environment successfully. TV based distance learning is a good example in this category. The student has only to go to a site and learn how to operate the audio button to ask a question. Technicians running the system ensure that the sending, receiving equipment and the cameras operate smoothly. The teacher does have to be trained in what can and cannot be presented in this medium and how to make effective use of cameras and audio channels. Similarly, environments have been created for web courses that integrate administration tools, content generation, and monitoring of students' progress; WEB-4M is a good example. CUSeeme (White Pine Software) is an environment that combines several features into one environment: video conferencing, whiteboard, e-mail, and document sharing among others.
IRI is an environment which is being developed with the goals of the ideal paradigm described in section 2.2 (a screen shot of the system in use can be seen in Figure 4). Currently this environment solves the scaling problem for up to an order of 100 users while maintaining tool-sharing capabilities by the use of reliable multicasting, specifically, RMP (Whetten). IP multicast does not support the reliable transport that is essential when transferring data, hence the integration of RMP. IRI, in its current state, does not support true interoperability (i.e. a user can be at either an NT computer or a Unix box). X tools can be shared to any platform but not PC tools. Because of its use of multicasting, IRI is totally symmetric. IRI supports both synchronous and asynchronous learning by recording live synchronous sessions in their totality (audio, video, and tool traffic) and automatically synthesizing the content for the asynchronous viewer, as described in Section 3. With the recording capability comes the possibility to take a course in a self-paced format.
The key to high perception and interactivity for IRI is its use of an Intranet rather than the Internet. In its current form it uses switched 10Mb/s Ethernet at each site, composed of an individual computer or a classroom of computers, and PVCs, with an average roundtrip time of 7ms. Although this limits the applicability of IRI, it is becoming quite prevalent to have Intranets with these capabilities. In the Hampton Roads region Cox Communications provides a relatively inexpensive service of 10Mb/s pipes to major sites and through the use of cable modems a 2.5Mb/s bi-directional pipe for the home user. Cox expects this kind of service to become prevalent in the near future in other regions as well. To connect sites in different regions, ATM service providers are also becoming prevalent. Judging from the regional experience, it is expected that the colocation factor should be extendable to cover the contiguous continental states. The extension of the scale will depend on advances in reliable multicasting; indications are that groups on the order of 1000 are within reach, through the use of hierarchical groupings. However, it should be clear that by having that many participants, class presentation will become more of a broadcasting model with interaction and tool use will be limited to a small percentage of the participating students.
IRI has developed sophisticated solutions to solve the problem of resource allocation management; resources are allocated to individual streams according to dynamic priorities. For example, when a heavy duty tool which uses most of the available network bandwidth, video quality is automatically reduced for all but the tool operator’s video. When certain clients have only the low end of bandwidth available then they are placed in a multicast group which receives lower quality video streams than the nodes on a high bandwidth network. The cost of using IRI is comparable to that of TV based distance learning systems because the workstation cost per student is amortized over many classes and general purpose computer lab use. Cost of PVCs for the wide area connectivity is already below that of leasing satellite channels, and Intranets do not need any incremental expenditures to support a learning environment. At the home, cable modem access will be priced in the same range as current Internet provider charges that most students already have.
This section focuses on the technical issues involved in building a web-based interface used to control an IRI session replaying previously recorded sessions. In the past, a Motif interface has been used to control an IRI session and add the resources (slides, etc) needed for a session. Since we are in the process of developing a cross-platform implementation of IRI, we decided to switch to a web-based controller as a first step. First we discuss how we enable multiple users to steer and monitor a session securely, and how we integrate their browsers with the IRI interface. In the second part, we present the process of content synthesis, and how such a content is presented to the user for arbitrary selection of replay.
IRI is based on reliable multicasting because of the number of students participating in a session. However, we expect only a few people, typically the teacher and perhaps an assistant, to be involved in the steering and monitoring process. Therefore, TCP/IP based, web server-to-browser communication can be used to coordinate the browsers belonging to the controller group.
From experience, it has been learned that a teacher is unable to both conduct a class and at the same time react to potential system problems. IRI is a complex distributed system and cannot react automatically to all possible faults; manual intervention is sometimes required. For instance, a public ATM network connects one of the sites in the current IRI system; this network is often overloaded, introducing delays of more than one second into the system. This delay causes time-out mechanisms in the reliable multicast protocol to trigger a failure signal, and human intervention is needed. Until IRI is totally fault tolerant, there will be a need for human controllers.
The entire interface is a set of CGI scripts/programs, JAVA scripts, and applets that access a protected directory on the server side. These scripts communicate with IRI through Unix sockets that have direct access to IRI files. Each web page presented to the user requires a proper authentication token. Authentication tokens are obtained through a home page, which validates the user’s Unix password and registration with IRI.
The process of starting an IRI session requires user authentication from the web interface displayed in Figure 1. The session is started from the same interface, after specifying, on subsequent web pages, the configuration for the session (machines to be used, servers, lesson plan requirements) and the location and identities of specified controllers and monitors (defaults are available for all of these). The web interface, pictured in Figure 2, is the means by which a teacher is able to save a session configuration, for future use. Through this, and subsequent applets, session specifications are made, such as: the tools to be used, and the tool mode, shared or local. At startup, IRI initiates browsers at pre-specified locations. These IRI-initiated browsers are given appropriate tokens, and are opened to the web page shown in Figure 1. The interface button for "class management" is used to connect to these browsers and make them active for all controllers and monitors.
At any time during a session, the teacher can click on "class management", bring the browser to the foreground, click on "start replay", and proceed as described below. Similarly, the monitor can bring up her web interface, and proceed to monitor the session, intervening the session, through the interface, if necessary.
Figure 1. IRI Control Panel
Figure 2. IRI Lesson Planning
This section describes how IRI can be used for both synchronous and asynchronous participation. IRI, as a software system, knows who is speaking, whose video is shown and what tools are running, at any point of time. The recording concept is simple: during a synchronous session, record all the individual streams and insert timing points. This information is synthesized and presented to the user as a set of web pages, which can be used at a later time to review any portion of the lecture through the web navigation pages. IRI can run an arbitrary X program, but clearly cannot go inside the program and deduce what events have occurred. For example, if during a lecture a student brings up a Netscape browser and visits various sites, IRI does not know which sites were visited; only the browser has that knowledge. In this case, IRI can create a page showing only that a particular student used the Netscape browser from time a until time b. IRI can still record the entire time and replay what was seen on the screen, but cannot provide more navigational help. However, there are a number of specialized tools, written especially for IRI, for which IRI tracks all events, and can synthesize a detailed sequence of these events. Examples are a slide show tool, a survey tool, and an exam tool. For example, a teacher may have used the slide show tool to display a number of slides from various presentations during a particular session. IRI can record each slide that was shown, and the time it was displayed, with respect to the beginning of the session. At the end of the lecture, IRI will synthesize web pages showing the titles of the displayed slides, and the time intervals they were shown. A single user, or a session, can then arbitrarily choose a starting point from this list of slides to begin replaying; the system will compute the video, audio, and tool streams that were active at the same time, and play them accordingly. In this approach there is no additional cost to provide the asynchronous view, while we maintain the ability to navigate through the material using web technology.
Assume that a group of students have started a session, and are reviewing a previously recorded session. The group leader started the session on his PC from his office, and has decided to steer the session remotely, from a machine in the IRI classroom. After logging on to the remote machine, he uses regular IRI features to communicate with the other students. At some point he brings his control browser into the foreground and clicks "activate replaying." A CGI script will ask him to identify what session to replay and when that is specified, present him with a synopsis of that session through JAVA applets, as shown in Figure 3. The JAVA applet knows already what session is to be replayed and presents a list of participants who spoke or had their image present and at what times of the session. In addition the leader can select the level of abstraction the presentations should be given. The choices are class, presentation, and slides. The first one implies that the applet will only show from which classes the leader of the recorded session used presentations. In the second choice, the applet will present the times each different presentation was on the screen during the session. In the example shown in Figure 3, the leader selected the third option and was given all the slides visited during the entire recorded session. The key is that each of these applets' windows can be left on the screen and at any time the leader can click on any of the time intervals to start replay at a particular point of time. These windows provide the synopsis of the session in different forms and depending on how a person remembers best, she can choose whatever stream is most appropriate.
Figure 3. Replay Control Panel
Once the leader clicks on a particular time interval, the applet computes, from the two shown endpoints of the interval, the time the user intended, and sends a signal to the replay IRI process. This process computes which IRI process was active at the selected time, and sends signals to activate the appropriate processes (audio, video, or presentation tool). The replaying process selects a window on the IRI interface, tags it as being a recorded window, and plays until the next event occurs, (e.g., stream ends or next slide is shown). In Figure 4 we show the result of the leader's action. The video windows on the right show two live students and one recorded teacher video. In the work area we see a live xterm and a recorded presentation which corresponds to the recorded teacher.
Figure 4. IRI Session with Live and Recorded Streams
At this point we have the first version of recording and replaying implemented and tested. The web interface for steering and monitoring is currently being developed, although not yet being used in an actual class environment. In the next phase, we will implement the same recording/replaying process for the remaining special IRI tools, such as survey, exam, and co-browsing. The next major step will be to synthesize and abstract the recording of arbitrary X tools. Since we have no control over these tools, we will not be able to provide any summary information except screen images at selected time points. We are investigating the concept of fast forward for the replaying of X tools, which will elide X commands which are not relevant after a certain time.
IRI is a complex distributed system that scales to the order of 100 simultaneous users, due to its use of reliable multicasting and efficient use of Unix communication mechanisms. We have shown that we can control this distributed system with a platform independent web interface. An important element is that the controller group is of an order of magnitude smaller than the number of learners participating in the session. IRI has a detailed knowledge about various multimedia and X streams; this can be used to provide sufficient information to JAVA applets. The applets then synthesize the content of a session, and present it to users who want to replay a previously recorded session, in its entirety, or from an arbitrary time within the recorded session.
Abdel-Wahab, H., and M. Feit (1991), Proceeding IEEE TriComm '91: Communications for Distributed Applications & Systems: Xtv: A Framework for Sharing X Window Clients in Remote Synchronous Collaboration, 159-167.
Davis, J. Z., K. Maly, and M. Zubair, A Coordinated Browsing System, NCSTRL as TR-97-29, Department of Computer Science, Old Dominion University, 1997.
Fox, E. and L. Kieffer. Multimedia curricula, courses and knowledge modules. ACM Computing Surveys, 27(4):549-551, Dec. 1995.
Maly, K., H. Abdel-Wahab, C. M. Overstreet, C. Wild, A. Gupta, A. Youssef, E. Stoica, and E. Al-Shaer (1997), IEEE Internet Computing: Interactive Distance Learning Over Intranets 1 1, 60-71.
Maly, K., C. Wild, C. M. Overstreet, H. Abdel-Wahab, A. Gupta, S. Purcell, A. Youssef, E. Stoica, R. Talla, and A. Prabhu (1996), J. of Networks and Computer Applications: Interactive Remote Instruction: Lesson Learned, 44-51.
Maly, K., C. Wild, C. M. Overstreet, H. Abdel-Wahab, A. Gupta, A. Youssef, E. Stoica, R. Talla, and A. Prabhu (1996), International J. of Innovations in Ed.: Virtual Classrooms and Interactive Remote Instruction, 44-51.
Whetten, B., T. Montgomery, and S. Kaplan, A high performance totally ordered multicast protocol, Theory and Practice in Distributed Systems, 1994.
JDH Technologies, Web-4M, www.jdhtech.com/web-4m.html.
Microsoft, NetMeeting, www.microsoft.com/products/prodref/113_ov.htm.
Network Research Group of the Information and Computing Science Division, vic and vat, www-nrg.ee.lbl.gov.
Oregon State University, QuestWriter, iq.orst.edu/doc/final/QWHome.html.
White Pine Software, CU-SeeMe, www.cu-seeme.com/cu31-win.html.