Hypertext Syllabi in Cognitive Science

Saul Traiger

Associate Professor

Cognitive Science Program

Occidental College

Los Angeles, CA 90041

traiger@oxy.edu

Telephone: (213) 259-2901

Hypertext Syllabi in Cognitive Science

I. Hypertext, Cognitive Science and Philosophy

Cognitive science is fertile ground for the study of on-line information delivery in academe. While the existence of most academic disciplines predate the computer, cognitive science owes its existence to it, embracing the power of the computer, both as a model of cognition, and as a tool in the delivery of the results of the investigation of such modeling. Every documentation delivery system makes assumptions about how individuals acquire information. Cognitive Science studies those assumptions directly, and so promises to inform the development of computer-based tools for the acquisition of information in all fields.

Cognitive science, however, is not a univocal field, but an interdisciplinary umbrella, under which fall at least philosophy, psychology, artificial intelligence, linguistics, and cognitive neuroscience. In this paper I will discuss hypertext applications in cognitive psychology and philosophy, the two fields in which I've developed such applications for use by undergraduates.

Philosophers have not ignored the new role of computers in higher education. Computers increase efficiency and provide new ways of conveying information. Most professional philosophers have computers on their desks which they use for writing papers, books, course materials, and what a colleague once called "interdepartmental ballistic memos." Electronic mail is increasingly available, and philosophers embrace it enthusiastically, not just as a replacement for interdepartmental mail, but for much of their professional work. The Hume Society, an international society which promotes scholarship on the Eighteenth Century philosopher David Hume, encourages members of its executive committee to master e-mail, since most decisions are made through an e-mail discussion group consisting of executive committee members. The American Philosophical Association maintains an electronic bulletin board for its members, and has just constructed a gopher server. (apa.oxy.edu)

The use of computers in philosophy courses has led to new questions about the discipline. Logic programs, for example, foster the development of techniques and algorithms in automated proof theory. (Ager, 1984), (Larsen, et. al., 1978) Programs such as Turing's World and Tarski's World enable students to construct Turing machines and semantics in first order logic, where the products of their efforts are not inert pieces of paper, but real machines. (Barwise, Etchemendy 1992) Computers are not just tools for organizing existing principles of philosophy and logic; rather their use can lead us to new understanding of those principles.

As a field contributing to cognitive science, the subject matter of philosophy has been changed by the computer even more radically. Philosophers must now come to grips with information which formerly appeared to be outside the scope of philosophy. A philosopher of mind, for example, in addition to mastering arguments for and against traditional theories, must come to understand the rudiments of computing machines. Further, the philosopher must look at ongoing research programs in such fields as psychology, neuroscience, and linguistics, where philosophical assumptions about the nature of cognition, language, and the mind in general are played out.

In psychology the influence of the computer is profound. It not only provides the theoretical underpinnings of much of cognitive psychology, (Pylyshyn, 1986) it has also changed the very nature of experimental psychology. One example of this is protocol analysis. Alan Newell and Herbert Simon pioneered the method by which the reports of experts are instantiated in computer programs, and then tested to see whether the computer exhibits expert behavior. (Newell and Simon, 1976)

How can we use computers to teach cognitive science effectively? Most academics are skilled at delivering information in traditional ways, through historical texts, journal articles, and the like, and these form the basis of our discussions and lectures. But they are likely to be less confident in the new areas mentioned above, where the need for new tools is critical. Of course we can draw on traditional texts, journal articles, and abstracts in fields like psychology and linguistics. This raises two questions: How do we "draw on" traditional resources? Are traditional resources the best ones to draw on? When we evaluate the possibility of alternative ways of presenting the subject matter of fields such as cognitive science, we need to understand something about the fields themselves.

Ted Nelson, who coined the term "hypertext", reports that he abandoned a book length work in philosophy because the constraints of normal text made it impossible for him to organize the complex relationships among concepts. (Nelson, 1992) I wouldn't want to speculate about the accuracy of Nelson's self-diagnosis. Certainly much excellent work has been done in philosophy before the advent of hypertext, and it doesn't follow from Nelson's philosophical writer's cramp using traditional text that hypertext is better suited to the domain. In fact, while I'll try to explain why hypertext is particularly well suited to the information-delivery needs of philosophers outside their usual domain, it can also be used in a more limited way for the kind of information which one needs to acquire in traditional philosophical subjects.

II. What is Hypertext?

There's little agreement on what hypertext is and perhaps less on what to call it. Some disagreement turns on the issue of scale: There are hypertext systems which link documents across networks and across continents, while others may simply link bits of text within the same document. Other disagreements concern the "text" part of "hypertext." Many hypertext systems, including the one I will describe below, can link not only bits of texts or documents which contain text, but graphic elements, programs, sound, and real-time video. Rather than quibble over terminology, I will simply list the most important properties of such systems, for which I'll use "hypertext" as a generic term. We still use the term 'word processor' for something which does more than process words.

A hypertext authoring tool combines the features of text and graphics-processing tools in a free-form database. A hypertext development system allows one to format text and graphics to reflect content in structures, much as an outline processor does, and it also facilitates dynamic cross-referencing and linking of text, graphics, programs, files, and other "objects". In a hypertext document text is dynamic in the sense that by selecting a bit of text one may access another bit of text with which the former is linked.

My interest in hypertext was stimulated by a demonstration of Brown University's Intermedia project. I.R.I.S., Brown's Institute for Research in Information and Scholarship developed a hypertext authoring system on IBM workstations, later on Apple computers running Unix. The project created courseware in a variety of subjects. (Landow 1992) I was particularly impressed with the ability of Intermedia to provide different access points to the same information. This is, in my view, something hypertext does extremely well. In history, for example, one might access the same information about the Civil War organized on a time line, by the names of important persons, by events, by political themes, and by geographical location. If students have many ways of organizing the same information, they are more likely to acquire that information, and they are more likely to be able to do something with the information, once acquired.

At the time Intermedia was in development at Brown, Occidental College, a small liberal arts college without a computer science major, owned one Sun workstation. We weren't about to launch a project with the hardware requirements of Intermedia. A search for an alternative authoring system began, with the chief requirement that it would run on IBM PCs or compatibles. By this time Hypercard for the Macintosh was available, but we found it to be too limited for our purposes. We turned instead to a product called Guide, from Owl International. Guide is based on Peter Brown's work on hypertext for Unix. (Brown, 1989) Versions for both Microsoft Windows 2.1 and the Macintosh were available. I set to work to see whether at least some of the functionality of Intermedia could be achieved on a more modest platform.

The hypertext syllabus I'll describe below was developed on several versions of Guide, including Guide 3.1, the most recent version of the software produced by InfoAccess, formerly Owl International. Guide has evolved with the evolution of Microsoft Windows, but it is a tribute to the early version that it contained most of the features of the current product. A Guide document contains text or graphics, and one can turn any text or graphic into one of several kinds of "buttons." The simplest type of button is the expansion button. Selecting a bit of text expands that text into more text. Expansion buttons work well for definitions. Click on a term and you get its definition. Reference buttons are text or graphic elements which are linked to some other text or graphic, either in the same document in another. Note buttons pop up the contents of text or graphics placed in a special definition window. Command buttons are the most powerful; they carry out a command, either one of several commands which execute one of several Guide interpreters, run programs written in Guide's C-like language Logiix or link to external DOS or Windows programs. Logiix programs are entered in the definition window of a document.

Guide has many options for controlling the format of documents, the behavior of the various buttons, and the overall appearance of the hypertext system. For example, while the Guide authoring system appears with menu, ribbon and command bars, those can easily be hidden, giving the author full control of the screen appearance. A reference button can be set up to close the document from which the reference was launched, when moving from one document to another; another option allows the document which contains the reference button to remain active while it calls another document. An expansion button can reveal text with or without hiding the expansion button itself. Guide keeps track of its actions, and has good backtracking capabilities. Unlike Intermedia, it does not create a web or map of the hypertext links in a document or set of documents.

Early versions of Guide had no provisions for the distribution of Guide-authored documents. One had to possess the development system in order to view Guide documents. Owl added first a low cost reader, and then a free "viewer". The viewer may be distributed royalty-free with guide documents.

III. The Hypertext Syllabus

Any course at a college or university can be viewed as a document delivery system. The documents include lectures and directed discussions, as well as books, articles, reference works, notes and diagrams. Sometimes these forms of information are accessed individually, as when one reads an assignment or listens to a lecture. But if one reflects on what happens in a good class, an exciting class, it's clear that many texts are invoked. The lecturer may quote an article with which the students are familiar from the previous night's reading, she may place notes up on a blackboard, and she will certainly weave her own verbal text. In classes like physics, psychology and linguistics, lectures may be peppered with demonstrations and experiments which are witnessed by students or in which students actively participate. The instructions for accessing these texts, and relating them, is the course syllabus.

The course syllabus points to the various texts, demonstrations, and videos which make up a course. The number of links or pointers to course materials contained in a syllabus will vary from syllabus to syllabus. Some syllabi specify reading and lecture content by the course meeting, others by the week, still others by theme. The point is that a good syllabus is a navigational tool for both the student and the instructor. It is a tool for document navigation, where documents are broadly construed. Thus it was only natural to use the concept of the course syllabus in designing a hypertext system which would organize the materials in cognitive science courses at Occidental College. The main idea is a simple one: I used Guide to create a hypertext syllabus which is simply a hypertext version of an existing syllabus.

I call one of my hypertext syllabi Hyperpsych, and I developed it for cognitive science students at Occidental College. Hyperpsych appears to the reader as a syllabus for a course in cognitive psychology. The differences, however, between Hyperpsych and a traditional syllabus are many. First, Hyperpsych appears in a window on the screen under Microsoft Windows, a popular graphical user interface (GUI). The reader or user controls the text by the use of a mouse. One simply "clicks" the mouse on any highlighted piece of text in the hypertext syllabus, and some action takes place. Clicking on "topics", for example, takes the user to a list of topics to be covered in the course. Choosing a topic takes the reader to a window which lists subtopics, the readings for that topic, and a "lecture notes" button. Clicking on the lecture notes button takes the reader to a new window, and closes the syllabus window. In the new window, an introduction to the topic is provided. From this window, the reader can explore the information in a particular area of cognitive psychology. At each point, one can easily back-track to the syllabus or some intermediary level.

Hyperpsych includes chapters on standard topics of cognitive psychology. We will concentrate on the chapter on problem solving. This is an area of great concern to philosophers interested in computers, and to those interested in the philosophy of psychology. Further, the "chapter" on problem solving demonstrates some of the special capabilities of hypertext. For example, it includes discussions of many different problems studies by cognitive psychologists, including the Tower of Hanoi, the Mutilated Checker Problem, chess and the Cannibals and Missionaries problem. Using Hyperpsych, the student can access information on these problems from a variety of perspectives: She can choose from a list of problems, from discussions of basic concepts, such as concepts like algorithm and heuristic, and from discussions of experiments. Perhaps most importantly, however, the student can explore the problems from within the hypertext system itself. She can click on text to reveal a window in which the Tower of Hanoi problem is executed by the computer, and then click on text to play chess or tick-tack-toe. Further, the hypertext system can display any graphical information, and Hyperpsych makes use of scanned images from a variety of sources.

As a brief demonstration reveals, Guide authored hypertext is extremely well suited to the task of accessing information on problem solving. It overcomes one of the biggest limitations of traditional textbooks: It is very difficult to understand the processes involved in problem solving without attempting to solve some of the classic problems studied by cognitive psychologists. Interaction with problems not only familiarizes one with the domain under investigation, but it often reveals some of the results. A classic problem studied in cognitive psychology is the Cannibals and Missionaries Problem. The problem is this: There are three cannibals and three missionaries on one side of a river. The task is to transport all six to the other side of river on a two person canoe while making sure that the cannibals never outnumber the missionaries on either bank. The Hyperpsych user is presented with a Cannibals and Missionaries game, in which she can load the canoe and transport passengers. The user is likely to get stuck at the point in the problem space where it appears that a critical move takes one away, rather than towards, the goal state. The experimental literature has shown that subjects usually get stuck at just this point. Students become, through interaction with the hypertext syllabus, the subjects of the experiments they are studying.

I also developed a second hypertext syllabus for a course in the philosophy of science. While the hypertext presentation of the material for this course facilitates the learning of technical terms and central concepts, it is not clear to me that it is as useful as the cognitive psychology hypertext. I argued earlier that hypertext may be more appropriate for one type of subject matter than another. The key is to look at how information is best presented in the subject, and then see whether a hypertext system can capture that mode of presentation. The crucial pieces of information in philosophy are arguments. Arguments include premises and a conclusion which is purported to follow from those premises. The multifarious links and cross-references achieved with hypertext don't capture the linear structure of an argument.

This doesn't mean that hypertext is of no use in philosophy or in other domains. Philosophical arguments are expressed in propositions. Propositions are composed of terms, often technical terms which must be understood by the reader of the argument. Hypertext is particularly useful for linking terms to definitions, or for linking terms to other contexts from which their meaning is more readily apparent. The future is particularly bright for the use of hypertext in critical editions of philosophical works. A critical edition must justify editorial choices from among the range of possibilities presented in various editions of that text. Textual variants and editorial explanations could easily be treated as hypertext, particularly in light of the fact that the critical editions of the future will be electronic texts first and foremost.

IV. Reading Comprehension with Hypertext

How does reading comprehension with hypertext compare with reading comprehension with traditional text, using the same subject matter? A study designed to answer this question was carried out in a senior comprehensive project I supervised at Occidental College by Indrani S. Stangl. (Stangl, 1991) Stangl created two versions of introductory material in linguistics, one in linear text form and another as hypertext. This subject matter was chosen in order not to favor hypertext. Stangl reasoned that if a hypertext with links to demonstrations and programs like the ones found in the problem solving module were used, hypertext would clearly win out. Stangl did not test that hypothesis however. Instead, using the material from an introduction to linguistics, she tested two groups for comprehension, one which used the hypertext and the other which read linear text. The hypertext readers were given a five minute instruction period. The sessions with hypertext were video-taped.

Stangl found that raw comprehension score was higher (58% correct) with linear text than with hypertext. (53% correct, t(33) = 1.00; n.s.) An analysis of these results and careful study of the videos suggests that hypertext readers were at a serious training disadvantage. Although the hypertext system was extremely easy to use and navigate, Stangl's five minute training session didn't begin to compensate for the difference in the ease of reading compared with students who read linear printed text. (Furnham and Barrie, 1989) This suggests that we need to factor in computer training time and the cost of such training when evaluating the overall effectiveness of on-line instruction. A positive result was that the correct answers of hypertext readers were more distributed over the subject matter than those of linear text readers, whose rate of correct answers fell off as questions moved from material from the beginning of the text to the end of the text. Thus hypertext readers had a better holistic comprehension. A study which compared a rich hypertext, such as the problem solving chapter in Hyperpsych, with linear text, might demonstrate the superiority of hypertext for comprehension.

V. Problems and Challenges

Several problems surfaced in the development of the hypertext syllabi. These range from specific hardware and software limitations to the question of the portability and use of hypertext systems. Making hypertext systems available to students is more costly than issuing textbooks. When one starts to consider the many types of costs, the issue is quite complicated. Even though the cost of computers and computer laboratory facilities continues to drop, making networked hypertext labs of the sort pioneered by I.R.I.S. much more widely than in the past, other costs will remain fixed and even escalate. One of these is the cost of maintaining software, networks, and workstations. Perhaps the most significant is the cost of developing hypertext documents in the first place. Colleges and universities are slow to change their reward structures. Authoring hypertext and other on-line documentation and education systems take significant efforts, efforts not well understood by administrators and colleagues, who have a handle on the effort required for the publication of books and articles, but not on those required to develop on-line documentation and training.

Another issue concerns access to computers. If we replace traditional texts with hypertexts, we need to make sure that the machines on which hypertexts run are readily available to students. Care must be taken to attend to the physical environment in which computers are placed. In an early installation, many students shared a single machine. While this placed a limit on the number of users, it created a cooperative learning environment. With better endowed labs, we should strive to make labs a place for interaction among students, and among students and instructors, as well as among students and machines. We can strive to enrich the computer-based learning environment, but it will be a long time before we can capture the richness of human-human interaction.

Two other large issues should be mentioned in closing. The first concerns matters of copyright and permissions. As Ted Nelson has argued, effective hypertext draws on materials from many authors. As Guide documents and others link out to the Internet and other networks and sources, we must begin to address cooperative schemes for document distribution. The second issue is that of standards for hypertext documents. There is much work in this area, and it cannot be surveyed here. The important point is that as hypertext competes with traditional texts, hypertext authors will require as stable a technology as is found in the traditional book.

References

Ager, T., (1984) "Computation in the philosophy curriculum", Computers and the Humanities, 18, pp. 145- 156.

Barrett, Edward, (1988) TExt ConText, and HyperText: Writing with and for the Computer Cambridge, MA: MIT Press.

Barwise, J. and Etchemendy, John, (1992) "The Language of First-Order Logic" Chicago: University of Chicago Press.

Brown, P.J. (1989) "A Hypertext System for UNIX.", Computing systems Wint 89 v 2 n 1.

Furnham, Adrian, and Barrie, Gunter (1989) "The Primacy of Print: Immediate Cued Recall of News as a Function of the Channel of Communication", The Journal of General Psychology, 116, pp. 305-310.

Landow, George P., (1992) Hypertext: The Convergence of Contemporary Critical Theory and Technology, Baltimore: The Johns Hopkins University Press.

Larsen I., Markosian L. and Suppes, P., (1978) "Performance models of undergraduate students on computer- assisted instruction in elementary logic", Instructional Science 7, pp. 15- 35.

Nelson, Theodor Holm, (1992) "Opening Hypertext: A Memoir", in Tuman, Myron C., ed., (1992) Literacy Online Pittsburgh: University of Pittsburgh Press. pp. 43-58.

Newell, Allen and Simon, Herbert, (1976) "Computer Science as Empirical Inquiry: Symbols and Search", Communications of the Association for Computing Machinery 19, March

Pylyshyn, Zenon (1986) Computation and Cognition, Cambridge: MIT Press

Stangl, Indrani (1991) "Reading Comprehension in Hypertext and Linear Text", Senior Thesis, Cognitive Science Program, Occidental College.