Other Considerations

There are a number of issues other than the incorporation of visualization applets into a hypertextbook of which a prospective author needs to be aware. A few are listed below:

• The hypertextbook should have an attractive, inviting portal (cover).

• The hypertextbook should appear to the learner as a single, integrated component, not a collec-tion of disorganized pieces.

• The hypertextbook should have a polished, professional look and feel that gives the learner confidence in its use. A consistent design theme and a consistent appearance to all of the pages and the linking structure are important.

• The use of distracting “bells and whistles” should be avoided.

• An organization of the material by way of hyperlinks that leads learners through the hypertext-book in different ways based on different learning levels or styles should be provided.

• Audio and video should be incorporated where it would be effective.

In other words, a hypertextbook should have a professional look and should make use of the unique capabilities offered through hyperlinking to reach out to students with varying learning needs.

3 Problems

No endeavor this large is without problems. One large hurdle is the amount of work needed to write a hypertextbook. The effort entailed in writing a traditional textbook is plenty already. The con-struction of a hypertextbook requires the additional work of providing an organization that caters to different learning needs, of (possibly) recording and integrating audio and video clips appropriately,

and of designing and including effective active learning applets. This additional work is substantial.

Consider just the development of visualization applets. Even passive learning applets, in which the learner participates only as a viewer, are time-consuming to design. Now consider further the effort involved when such applets must be constructed to support active learning, to be adaptable to various situations, to be integrated into a hypertextbook, to interact with different but related applets, and to come supplied with authoring tools. It should be clear that such applets require substantial effort to design and implement well.

4 An Example

Let’s consider an example. The figure below is taken from a mockup of a portion of a hypertextbook on the theory of computing.

Even though the figure is small, there are a number of points to notice.

• The window that frames the entire figure is a standard browser window—the learner in this ex-ample has accessed the hypertextbook simply by linking to the hypertextbook through a browser in platform independent fashion.

• There is text at the top of the window and a finite state automaton applet at the bottom. The text includes a description of an exercise that is to be carried out with the finite state automaton ap-plet. The finite state automaton applet is embedded directly in the text where it should naturally appear and is not required to be opened in a separate window.

• The finite state automaton applet is a fully self-contained visualization system for finite state automata.

– There is a menu bar at the beginning of the applet with menu items for running a finite state automaton, clearing the finite state automaton pane (to allow construction of a finite state automaton from scratch), selecting an alphabet over which a constructed finite state automaton is to operate, and comparing a constructed finite state automaton against a correct solution.

– There is a tape for input (the linear sequence of cells below the menu and above the finite state automaton pane). The learner can type an arbitrary string onto this tape and then run the automaton in the pane below on that tape.

– There is a large pane (the one that displays the automaton of this example) in which a finite state automaton (either deterministic or nondeterministic) can be constructed and modified quite easily through a set of mouse clicks and drags.

– There is a set of “processing indicators” to the right of the automaton pane that display the status of the finite state automaton as it is being run on an input string (including acceptance or rejection of the input string).

• The learner in this example has apparently constructed a finite state automaton in the automaton pane in completing the exercise described in the text in which the applet is embedded. To check whether the constructed solution is correct, the learner has also apparently clicked the Compare button on the menu. The applet has returned a message saying that the automaton is correct. Had the automaton not been correct the message returned would have stated this fact and provided a sample string that the learner’s automaton accepts that is not in the language of the exercise, or a string that the learner’s automaton does not accept that is in the language of the exercise, thus providing guidance to help lead the learner to a correct solution.

The above observations from just this one example should help one gain an appreciation for just how much time must be invested in designing good active learning applets for inclusion in a hyper-textbook. Way back in 1974 in The Mythical Man MonthFred Brooks observed that if one were to design a software system for personal use that took timeNto complete, then if that same system were to be designed for use by others, or if it were to be designed to be integrated into some other existing system it would take time 3N to do. If instead the original system were to be designed to be used by othersandto be included in some other existing system it would take time 9Nto complete. We have certainly informally validated this observation during our attempts to design software like the finite state automaton applet illustrated above, which is expressly intended to be used by novices and which also must be integrated into a hypertextbook and be able to interact seamlessly with other applets, such as a regular expression applet and a regular grammar applet.

The time factor for constructing useful applets is further increased when one wants to incorporate active learning. The integration of guidance and feedback mechanisms is a challenging task. In the illustration above, for example, the determination of whether a learner has constructed a finite state automaton that meets the specifications of an exercise requires the following:

• The person designing the exercise must provide a correct finite state automaton for the exercise that is hidden from the learner and which the applet must be able to maintain somewhere.

• The learner’s attempt at creating a correct finite state automaton must be compared with the correct, hidden finite state automaton. The comparison cannot be done in the applet by com-paring the two automata, because there are in general an infinite number of finite state automata recognizing any given language, so somehow the applet must determine whether thelanguages recognized by the two automata are the same.

• Using known results from the theory of computing, a finite state automatonM₁is be constructed from the learner’s automaton such thatM₁ recognizes the language that is the intersection of the language recognized by the learner’s automaton,ML, and thecomplementof the language recognized by the correct automaton,M_C. If this new automaton,M₁, accepts any strings at all,

this means that the learner’s automaton, ML, accepts some strings that the correct automaton M_C does not accept. Examples of such strings can be found by doing a depth-first search on the graph for M₁, which can be presented to the learner for feedback. Similarly, a finite state automatonM₂ can be constructed that recognizes the language that is the intersection of thecomplementof the language recognized by the learner’s automaton,ML, and the language recognized by the correct automaton,M_C. IfM₂accepts any strings at all, then there are strings that the correct automaton accepts that the learner’s automaton does not accept. Again, a depth first search on the automatonM₂will reveal examples of such strings for feedback to the user.

If bothM₁ andM₂ recognize the empty language, then the learner has constructed a proper automaton, that is, one that recognizes the same language as the hidden correct automaton.

Since active learning and feedback are important objectives of applets used for visualizing edu-cational concepts, it should be clear from this description that the applets are going to necessarily be more complex and time-consuming to construct than similar passive learning applets.

5 Evaluation

Do hypertextbooks work? Phrased this way, the question is perhaps too broad. More specifically, we want to know the answers to some of the following questions:

• Do students learn better in traditional learning environments with hypertextbooks that incorpo-rate active learning modules?

• Do students learn faster in traditional learning environments with such hypertextbooks?

• Can students learn better and/or faster in non-traditional learning environments (e.g., in distance learning situations) when hypertextbooks that incorporate active learning modules are available for the subject?

• Do students perceive that their opportunities to learn are enhanced when such hypertextbooks are used?

• Do students enjoy learning more when such hypertextbooks are used as opposed to traditional textbooks?

We would be happy to determine, for example, that students learned at least as well with hypertext-books as with traditional texthypertext-books if at the same time they were more satisfied with the learning process. This might encourage better learning and even entice more underrepresented classes of stu-dents (e.g., women) to complete a degree in computer science.

Our hypertextbook projects have only recently begun to undergo preliminary evaluation. The thesis cited in the References section below describes some of these evaluations.

6 Summary

In this paper, we have presented an outline of our investigations into the feasibility and effectiveness of hypertextbooks, particularly those that incorporate active learning visualization applets. Our ongoing work can be followed at

http://www.cs.montana.edu/webworks 7 References

Rather than providing a small list of references that would fit in this paper, we encourage readers to obtain a copy of the thesis Active Learning Animations for the Theory of Computation by the first author. The thesis contains a much more detailed look at the construction of active learning applets and includes an extensive list of references. The thesis can be obtained online at

http://www.cs.montana.edu/webworks and at

http://www.cs.mtech/~grinder

The authors can also be contacted directly for information about the hypertextbook projects.

Ben-Bassat Levy R., 83